Practice Requirements Forecasts Requirements Hydrological Forecasts Forecasts Forecasts Hydrological Forecasts and Best Practice Practice Requirements Requirements Hydrological Best Practice Hydrological Best Practice ForecastsRequirements Forecasts Best Practice Forecasts Requirements Best Forecasts Requirements Best Hydrological Requirements and Hydrological Forecasts and Hydrological Best Hydrological Forecasts Best Best Practice Practice Forecasts Best Practice Hydrological Practice Best Requirements Forecasts Hydrological Forecasts Forecasts Best Requirements Best PracticeHydrological and Practice Forecasts Best Practice Forecasts Best Practice Requirements Hydrological Best Hydrological Practice Best Training Module

Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practices

A. Vogelbacher PracticePractice Requirements Forecasts Requirements Hydrological Forecasts Forecasts Forecasts Hydrological Forecasts and Best Practice Practice Requirements Requirements Hydrological Best Practice Hydrological Best Practice ForecastsRequirements Forecasts Best Practice Forecasts Requirements Best Forecasts Requirements Best Hydrological Requirements and Hydrological Forecasts

and Hydrological Best Hydrological Forecasts Best Best Practice Practice Forecasts Best Practice Training Module Hydrological Practice Best Requirements Forecasts Requirements Forecasts HydrologicalFlood Disaster Forecasts Risk Management- Best Best PracticeHydrologicalHydrological Forecasts: Requirements andand Practice Forecasts BestBest Practice PracticesForecasts Best Practice Requirements Hydrological Best Hydrological Practice Best PracticePractice Requirements Forecasts Requirements Hydrological Forecasts Forecasts Forecasts Hydrological Forecasts and Best Practice Practice Requirements Requirements Hydrological Best Practice Hydrological Best Practice ForecastsRequirements Forecasts Best Practice Forecasts Requirements Best Forecasts Requirements Best Hydrological Requirements and Hydrological Forecasts and Hydrological Best Hydrological Forecasts Best Best Practice Practice Forecasts Best Practice Training Module Hydrological Practice Best Requirements Forecasts Requirements Forecasts HydrologicalFlood Disaster Forecasts Risk Management- Best Best PracticeHydrologicalHydrological Forecasts: Requirements andand Practice Forecasts BestBest Practice PracticesForecasts Best Practice Requirements Hydrological Best Hydrological Practice Best Message

ISBN: 978-3-944152-15-8 Since 2010, GIZ has been collaborating with the National ©NIDM & GIZ, 2013 Institute of Disaster Management for implementing the Published by “Environmental Knowledge and Disaster Risk Management” National Institute of Disaster Management (NIDM), project, aimed at strengthening capacity building initiatives in Ministry of Home Affairs knowledge management and risk reduction for disasters caused 5-B, IIPA Campus, IP Estate, Mahatma Gandhi Marg by natural hazards, such as , cyclones, drought, or New Delhi 110 002, India T: +91 11 23702432, 23705583, 23766146 manmade disasters caused by industry. The design and F: +91 11 23702442, 23702446 development of training tools, such as an internet based training I: www.nidm.gov.in and knowledge management system and blended learning and training methodology and the development of training materials Deutsche Gesellschaft für are important activities under this project. Internationale Zusammenarbeit (GIZ) GmbH Indo-German Environment Partnership It gives me great pleasure to introduce the Training Module, Hydrological Forecasts - Requirements and Best Practices. I hope B-5/2 Safdarjung Enclave Dr. Dieter Mutz New Delhi 110 029, India the case study will help trainees in developing better Director T: +91 11 49495353 understanding of the early warning, forecasting and F: +91 11 49495391 Indo-German Environment dissemination mechanisms. I: www.giz.de Partnership (IGEP) Author Programme Deutsche I take this opportunity to express appreciation of the commitment Gesellschaft für A. Vogelbacher, Director, Flood Information Centre, Bavarian Environment Agency, Germany of NIDM, Ministry of Home Affairs, Government of India, New Internationale Delhi, the Flood Information Centre, Bavarian Environment Review and Editing Zusammenarbeit (GIZ) Agency, Munich, Germany and ifanos Germany and India who Anil K. Gupta, Associate Professor, NIDM GmbH made this effort successful. I wish that such modules are used Sreeja S. Nair, Assistant Professor, NIDM New Delhi, February 2013 Florian Bemmerlein-Lux, Director, Ifanos Germany & Consultant GIZ extensively by all stakeholders across the country.

Acknowledgements Dr. Satendra, IFS, Executive Director, NIDM Dr. Dieter Mutz, Director, IGEP, SunandaDey, Delhi University, Delhi Citation: Vogelbacher, A. (2013). Flood Disaster Risk Management - Hydrological Forecasts: Requirements and Best Practices (Training Module). National Institute of Disaster Management, New Delhi and Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) GmbH, Germany, 88 p.

Design and Printing M/s Rouge Communications, S-185, Greater Kailash Part 2, New Delhi, February, 2013

Disclaimer This document may be freely reviewed, reproduced or translated, in part or whole, purely on non-profit basis for any non- commercial and academic purpose aimed at training of education promotion as cause for disaster risk management and emergency response, keeping the source acknowledged. Authors welcome suggestions on its use in actual training situations and for improved future editions. The present document is neither exhaustive nor complete on the topic of hydrological forecasts. The information has been compiled from reliable documents and published references/resources, as cited in the publication. Mention of any company, association or product in this document is for informational purpose only and does not constitute a recommendation of any sort by either NIDM or GIZ.

(ii) (iii) Message

ISBN: 978-3-944152-15-8 Since 2010, GIZ has been collaborating with the National ©NIDM & GIZ, 2013 Institute of Disaster Management for implementing the Published by “Environmental Knowledge and Disaster Risk Management” National Institute of Disaster Management (NIDM), project, aimed at strengthening capacity building initiatives in Ministry of Home Affairs knowledge management and risk reduction for disasters caused 5-B, IIPA Campus, IP Estate, Mahatma Gandhi Marg by natural hazards, such as floods, cyclones, drought, or New Delhi 110 002, India T: +91 11 23702432, 23705583, 23766146 manmade disasters caused by industry. The design and F: +91 11 23702442, 23702446 development of training tools, such as an internet based training I: www.nidm.gov.in and knowledge management system and blended learning and training methodology and the development of training materials Deutsche Gesellschaft für are important activities under this project. Internationale Zusammenarbeit (GIZ) GmbH Indo-German Environment Partnership It gives me great pleasure to introduce the Training Module, Hydrological Forecasts - Requirements and Best Practices. I hope B-5/2 Safdarjung Enclave Dr. Dieter Mutz New Delhi 110 029, India the case study will help trainees in developing better Director T: +91 11 49495353 understanding of the flood early warning, forecasting and F: +91 11 49495391 Indo-German Environment dissemination mechanisms. I: www.giz.de Partnership (IGEP) Author Programme Deutsche I take this opportunity to express appreciation of the commitment Gesellschaft für A. Vogelbacher, Director, Flood Information Centre, Bavarian Environment Agency, Germany of NIDM, Ministry of Home Affairs, Government of India, New Internationale Delhi, the Flood Information Centre, Bavarian Environment Review and Editing Zusammenarbeit (GIZ) Agency, Munich, Germany and ifanos Germany and India who Anil K. Gupta, Associate Professor, NIDM GmbH made this effort successful. I wish that such modules are used Sreeja S. Nair, Assistant Professor, NIDM New Delhi, February 2013 Florian Bemmerlein-Lux, Director, Ifanos Germany & Consultant GIZ extensively by all stakeholders across the country.

Acknowledgements Dr. Satendra, IFS, Executive Director, NIDM Dr. Dieter Mutz, Director, IGEP, SunandaDey, Delhi University, Delhi Citation: Vogelbacher, A. (2012). Flood Disaster Risk Management - Hydrological Forecasts: Requirements and Best Practices (Training Module). National Institute of Disaster Management, New Delhi and Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) GmbH, Germany, 88 p.

Design and Printing M/s Rouge Communications, S-185, Greater Kailash Part 2, New Delhi, February, 2013

Disclaimer This document may be freely reviewed, reproduced or translated, in part or whole, purely on non-profit basis for any non- commercial and academic purpose aimed at training of education promotion as cause for disaster risk management and emergency response, keeping the source acknowledged. Authors welcome suggestions on its use in actual training situations and for improved future editions. The present document is neither exhaustive nor complete on the topic of hydrological forecasts. The information has been compiled from reliable documents and published references/resources, as cited in the publication. Mention of any company, association or product in this document is for informational purpose only and does not constitute a recommendation of any sort by either NIDM or GIZ.

(ii) (iii) Foreword

Knowledge of environmental systems and processes are key factors in the management of disasters, particularly the hydro- meteorological ones. Climate-change is the challenge of modern times known to aggravate natural hazards like floods, drought, cyclone, landslides and forest fires, and on the other hand it also intensifies people’s vulnerability by affecting their resources and capacities. Environmental conditions including climatic and topographic factors also determine the dispersion, transport and thereby, the fate of chemical incidences.

NIDM and GIZ Germany, under the aegis of Indo-German Environment Partnership Programme (IGEP) with Indian Ministry of Environment and Forests, implemented a joint project entitled “Environmental Knowledge for Disaster Risk Management Dr. Satendra, IFS, (ekDRM)” wherein development of case studies and training Executive Director, modules are the key activities. NIDM New Delhi, February 2013 I am pleased that the five modules are developed under the project viz.

¢ Environmental Legislation for Disaster Risk Management ¢ Disaster Risk Management using Knowledge - Base and Statistics ¢ Flood Disaster Risk Management - Gorakhpur Case Study Module, ¢ Hydrological Forecasts - Requirements and Best Practices: Case Study Module ¢ Critical Infrastructures and Disaster Risk Reduction in the Light of Natural Hazards

The Case Study Module Hydrological Forecasts - Requirements and Best Practices demonstrates with example of operational flood forecasting, flood warning and dissemination systems in Bavaria and its comparison with India.

The efforts made by the author are commendable and I hope the module shall be used by trainers, scholars and practitioners at different levels. I am sure, the modules shall be of significant contribution in the training and capacity building programmes within the country and outside. I welcome the views and suggestions of the readers for improvements in future.

(v) Foreword

Knowledge of environmental systems and processes are key factors in the management of disasters, particularly the hydro- meteorological ones. Climate-change is the challenge of modern times known to aggravate natural hazards like floods, drought, cyclone, landslides and forest fires, and on the other hand it also intensifies people’s vulnerability by affecting their resources and capacities. Environmental conditions including climatic and topographic factors also determine the dispersion, transport and thereby, the fate of chemical incidences.

NIDM and GIZ Germany, under the aegis of Indo-German Environment Partnership Programme (IGEP) with Indian Ministry of Environment and Forests, implemented a joint project entitled “Environmental Knowledge for Disaster Risk Management Dr. Satendra, IFS, (ekDRM)” wherein development of case studies and training Executive Director, modules are the key activities. NIDM New Delhi, February 2013 I am pleased that the five modules are developed under the project viz.

¢ Environmental Legislation for Disaster Risk Management ¢ Disaster Risk Management using Knowledge - Base and Statistics ¢ Flood Disaster Risk Management - Gorakhpur Case Study Module, ¢ Hydrological Forecasts - Requirements and Best Practices: Case Study Module ¢ Critical Infrastructures and Disaster Risk Reduction in the Light of Natural Hazards

The Case Study Module Hydrological Forecasts - Requirements and Best Practices demonstrates with example of operational flood forecasting, flood warning and dissemination systems in Bavaria and its comparison with India.

The efforts made by the author are commendable and I hope the module shall be used by trainers, scholars and practitioners at different levels. I am sure, the modules shall be of significant contribution in the training and capacity building programmes within the country and outside. I welcome the views and suggestions of the readers for improvements in future.

(v) Contents

1 Introduction 1 8 Learning Unit - G: Updating and Assimilation 51 8.1 Updating of input variables 52 2 Learning Unit - A: Floods – Hydrologic Basics 3 8.2 Updating of state variables 52 2.1 Rain 4 8.3 Updating of model parameters 53 2.2 Influences of land surface 9 8.4 Updating of output variables or error prediction 53 2.3 and runoff concentration 9 2.4 Snow melt 10 9 Learning Unit - H: Quality Assessment and Uncertainty in Forecast 55 2.5 flow and flood routing 11 10 Learning Unit - I: Organization and Operation of Forecast 59

3 Learning Unit - B: Flood Warning Systems 13 10.1 Operating procedures 60 3.1 Bavarian flood information service 15 10.2 Staff 62

3.2 Water level and flood risk in Bavaria 17 Further Information 65 3.3 Website of the Bavarian Flood Information Service 18 Glossary 67 4 Learning Unit - C: Real Time Data Acquisition and Monitoring Network 21 4.1 Satellite-based estimates of rainfall and soil moisture 24 Bibliography 69

4.2 Flood related weather observation in India 26 About NIDM 71

5 Learning Unit - D: Meteorological Forecasts 29 About GIZ 72 5.1 German Weather Service (DWD) 30 About IGEP 73 5.2 Numerical weather prediction in India 31

About ekDRM 74 6 Learning Unit - E: Flood Forecast System 33

6.1 Hydrologic response time and lead time of the forecast 34 About the Author 75

7 Learning Unit - F: Hydrologic and Hydraulic Models 37 7.1 Black box model 38 7.2 Distributed models 40 7.3 The LARSIM model, as an example for a distributed conceptual model 44 7.4 Hydraulic models 49 7.5 The trend from 1 to 2-dimensional modelling 49

(vi) (vii) Contents

1 Introduction 1 8 Learning Unit - G: Updating and Assimilation 51 8.1 Updating of input variables 52 2 Learning Unit - A: Floods – Hydrologic Basics 3 8.2 Updating of state variables 52 2.1 Rain 4 8.3 Updating of model parameters 53 2.2 Influences of land surface 9 8.4 Updating of output variables or error prediction 53 2.3 Surface runoff and runoff concentration 9 2.4 Snow melt 10 9 Learning Unit - H: Quality Assessment and Uncertainty in Forecast 55 2.5 Channel flow and flood routing 11 10 Learning Unit - I: Organization and Operation of Forecast 59

3 Learning Unit - B: Flood Warning Systems 13 10.1 Operating procedures 60 3.1 Bavarian flood information service 15 10.2 Staff 62

3.2 Water level and flood risk in Bavaria 17 Further Information 65 3.3 Website of the Bavarian Flood Information Service 18 Glossary 67 4 Learning Unit - C: Real Time Data Acquisition and Monitoring Network 21 4.1 Satellite-based estimates of rainfall and soil moisture 24 Bibliography 69

4.2 Flood related weather observation in India 26 About NIDM 71

5 Learning Unit - D: Meteorological Forecasts 29 About GIZ 72 5.1 German Weather Service (DWD) 30 About IGEP 73 5.2 Numerical weather prediction in India 31

About ekDRM 74 6 Learning Unit - E: Flood Forecast System 33

6.1 Hydrologic response time and lead time of the forecast 34 About the Author 75

7 Learning Unit - F: Hydrologic and Hydraulic Models 37 7.1 Black box model 38 7.2 Distributed models 40 7.3 The LARSIM model, as an example for a distributed conceptual model 44 7.4 Hydraulic models 49 7.5 The trend from 1 to 2-dimensional modelling 49

(vi) (vii) Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best Flood hazards cannot be avoided and floods cannot be controlled fully This Case Study explores flood forecasting systems from the perspective of its position within the flood warning by engineering measures. But the mitigation of the consequences can process. A method for classifying the different approaches taken in flood forecasting is introduced before the help significantly reduce the risk posed. Prerequisite of any effective elements of a present-day flood forecasting system are discussed in detail. Finally, the state of the art in mitigation of the consequences of flooding is a good understanding of developing flood forecasting systems is addressed including how to deal with specific challenges posed. the development of flood events and subsequently the provision of effective flood warning. The target group of this case study are decision makers in disaster risk management and/or water management. The case study should help to understand some hydrologic basics of the flood forecast and Operational flood warning systems that provide timely warning will assist in the administration and implementation of an appropriate flood warning system in a specific reduce loss of life and property. Functioning warning systems are the environment, to find the best solution for a region. 1 basis to implement simple, yet effective measures such as evacuation, temporary relocation, or implementation of a Best solutions depend mainly on quality and availability of data, the areas and/or points of interest, catchment Introduction strategy. Additionally such systems are a cost-effective way of properties, cross border catchments, and financial capabilities with special consideration of flood forecast. reducing flood risk and may help avoid large investments in traditional engineering flood control measures such as the raising of dykes or the building of flood Control .

However, early warning is not sufficient. It must be always complemented by ensuring that the information reaches the vulnerable on time. It requires the awareness and the preparedness of those to be informed and those that have the task to warn.

Operational flood warning systems have been developed or are under development in basins worldwide. They all rely on the detection of floods through hydro-meteorological observation networks. The use of observation data is a primary element of a flood warning system. To increase the potential utility of the flood warning service through extension of the lead time of the prediction of a flood event, the state- of-the-art systems also incorporate some form of (model based) flood forecasting.

A wide range of techniques may be employed to provide flood forecasting capabilities, ranging from data-driven modelling techniques, through conceptual modelling approaches to more complex systems of conceptual and physical models (Werner et al., 2006).

1 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice Introduction 2 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best Flood hazards cannot be avoided and floods cannot be controlled fully This Case Study explores flood forecasting systems from the perspective of its position within the flood warning by engineering measures. But the mitigation of the consequences can process. A method for classifying the different approaches taken in flood forecasting is introduced before the help significantly reduce the risk posed. Prerequisite of any effective elements of a present-day flood forecasting system are discussed in detail. Finally, the state of the art in mitigation of the consequences of flooding is a good understanding of developing flood forecasting systems is addressed including how to deal with specific challenges posed. the development of flood events and subsequently the provision of effective flood warning. The target group of this case study are decision makers in disaster risk management and/or water management. The case study should help to understand some hydrologic basics of the flood forecast and Operational flood warning systems that provide timely warning will assist in the administration and implementation of an appropriate flood warning system in a specific reduce loss of life and property. Functioning warning systems are the environment, to find the best solution for a region. 1 basis to implement simple, yet effective measures such as evacuation, temporary relocation, or implementation of a flood control Best solutions depend mainly on quality and availability of data, the areas and/or points of interest, catchment Introduction strategy. Additionally such systems are a cost-effective way of properties, cross border catchments, and financial capabilities with special consideration of flood forecast. reducing flood risk and may help avoid large investments in traditional engineering flood control measures such as the raising of dykes or the building of flood Control dams.

However, early warning is not sufficient. It must be always complemented by ensuring that the information reaches the vulnerable on time. It requires the awareness and the preparedness of those to be informed and those that have the task to warn.

Operational flood warning systems have been developed or are under development in river basins worldwide. They all rely on the detection of floods through hydro-meteorological observation networks. The use of observation data is a primary element of a flood warning system. To increase the potential utility of the flood warning service through extension of the lead time of the prediction of a flood event, the state- of-the-art systems also incorporate some form of (model based) flood forecasting.

A wide range of techniques may be employed to provide flood forecasting capabilities, ranging from data-driven modelling techniques, through conceptual modelling approaches to more complex systems of conceptual and physical models (Werner et al., 2006).

1 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice Introduction 2 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best The most common cause of floods is intense rainfall. Flood occurs when it rains heavily and runoff is strong. Runoff is the most important component for flood prediction. Most dangerous flooding rainfalls are connected to certain weather conditions. For the extent and progression of a flood the percentage of the catchment area of a river affected simultaneously by rainfall plays an important role. A short, small-scale thunder storm can make an individual mountain into a raging white water, while the water level of a major river downstream fluctuates hardly. Only a steady rain that lasts long and 2 covers a large area can overtop the banks of large and . In the mountains melting snow can generate a large amount of LU-A: runoff. The faster and more snow melts, the greater the burden on the rivers. Ice can impede the flow of water volumes.

Floods- Runoff is the portion of rain and snow melt water that moves toward a stream channel rather than infiltrating the soil. For some purposes, Hydrologic however, runoff also includes the subsurface water known as inter flow which also quickly moves toward the stream channel. Figure 2.1: Main hydrologic processes in the water cycle (Mauser, 2012)

Basics Mainly there are two flood types taken into consideration in this study: 2.1 Rain ¢ Flash Floods For the generation of floods two types of rain are of particular importance: prolonged large-scale rainfall and ¢ Riverine Floods (slow river floods) short but intense duration rainfall accompanied by thunder-storm. A is a flood of short duration with relatively high peak . It is a flood that rises and falls quite rapidly with little or no Long lasting and wide spread rains have special influence on the extent of great river floods. This advective or cyclonic type of rain results from the convergence and subsequent lifting of air masses (cooling, condensation advance warning. The triggering event can be intense rainfall, the and growth of water droplets) in connection to a low-pressure area (cyclonic cooling). This type of storm may failure of a , , or other structure that is impounding water or last from 24 to 72 hours and deliver a total of 50 to 150 mm. Tropical rains of this type may reach up to the sudden rise of water level associated with river ice jams. As a 350 mm in a period of 12 to 24 hours. fluvial flood, a flash flood is limited to relatively small catchments.

The frontal type of lifting occurs when the circulation forces air up over a frontal surface. Inflowing warm air Riverine floods are medium or slow onset floods, which take a longer can be lifted at hitting colder air (warm front). This results in moderate rainfall rates often of quite long time to develop. This type of flood is connected to larger rivers and duration. A cold front forces a rapid raising and cooling of the displaced warm air. The resulting weather may streams with larger catchments. be tumultuous, with short-duration pelting rains and high winds.

3 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-A: Floods - Hydrologic Basics 4 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best The most common cause of floods is intense rainfall. Flood occurs when it rains heavily and runoff is strong. Runoff is the most important component for flood prediction. Most dangerous flooding rainfalls are connected to certain weather conditions. For the extent and progression of a flood the percentage of the catchment area of a river affected simultaneously by rainfall plays an important role. A short, small-scale thunder storm can make an individual mountain stream into a raging white water, while the water level of a major river downstream fluctuates hardly. Only a steady rain that lasts long and 2 covers a large area can overtop the banks of large rivers and streams. In the mountains melting snow can generate a large amount of LU-A: runoff. The faster and more snow melts, the greater the burden on the rivers. Ice can impede the flow of water volumes.

Floods- Runoff is the portion of rain and snow melt water that moves toward a stream channel rather than infiltrating the soil. For some purposes, Hydrologic however, runoff also includes the subsurface water known as inter flow which also quickly moves toward the stream channel. Figure 2.1: Main hydrologic processes in the water cycle (Mauser, 2012)

Basics Mainly there are two flood types taken into consideration in this study: 2.1 Rain ¢ Flash Floods For the generation of floods two types of rain are of particular importance: prolonged large-scale rainfall and ¢ Riverine Floods (slow river floods) short but intense duration rainfall accompanied by thunder-storm. A flash flood is a flood of short duration with relatively high peak discharge. It is a flood that rises and falls quite rapidly with little or no Long lasting and wide spread rains have special influence on the extent of great river floods. This advective or cyclonic type of rain results from the convergence and subsequent lifting of air masses (cooling, condensation advance warning. The triggering event can be intense rainfall, the and growth of water droplets) in connection to a low-pressure area (cyclonic cooling). This type of storm may failure of a dam, levee, or other structure that is impounding water or last from 24 to 72 hours and deliver a total of 50 to 150 mm. Tropical rains of this type may reach up to the sudden rise of water level associated with river ice jams. As a 350 mm in a period of 12 to 24 hours. fluvial flood, a flash flood is limited to relatively small catchments.

The frontal type of lifting occurs when the circulation forces air up over a frontal surface. Inflowing warm air Riverine floods are medium or slow onset floods, which take a longer can be lifted at hitting colder air (warm front). This results in moderate rainfall rates often of quite long time to develop. This type of flood is connected to larger rivers and duration. A cold front forces a rapid raising and cooling of the displaced warm air. The resulting weather may streams with larger catchments. be tumultuous, with short-duration pelting rains and high winds.

3 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-A: Floods - Hydrologic Basics 4 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best When moist winds blow up a slope, the air will expand and cool at the lower pressure corresponding to the higher elevation (orographic cooling).

Figure 2.3: Orographic lifting (MetEd, 2012)

Convection currents (convective cooling) develop under the condition of vertical instability in the atmosphere. If the moist air temperature decreases by more than 6ºC per kilometer it indicates unstable condition. This instability can be caused by surface heating. The air rises because it is warmer and less dense. The air is lifted to its level of free convection (LFC) and continues to rise as long as it is warmer than the surrounding air. Eventually, the rising air reaches an equilibrium level at which its temperature equals that of the surrounding Figure 2.2: Cold front & warm front (MetEd, 2012) environment (Figure 2.4).

5 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-A: Floods - Hydrologic Basics 6 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best When moist winds blow up a slope, the air will expand and cool at the lower pressure corresponding to the higher elevation (orographic cooling).

Figure 2.3: Orographic lifting (MetEd, 2012)

Convection currents (convective cooling) develop under the condition of vertical instability in the atmosphere. If the moist air temperature decreases by more than 6ºC per kilometer it indicates unstable condition. This instability can be caused by surface heating. The air rises because it is warmer and less dense. The air is lifted to its level of free convection (LFC) and continues to rise as long as it is warmer than the surrounding air. Eventually, the rising air reaches an equilibrium level at which its temperature equals that of the surrounding Figure 2.2: Cold front & warm front (MetEd, 2012) environment (Figure 2.4).

5 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-A: Floods - Hydrologic Basics 6 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best The southwest monsoon is generally expected to begin around the start of June and fade down by the end of

12 km 200 mb September. The moisture-laden winds on reaching the southernmost point of the Indian Peninsula, due to its 11 topography, become divided into two parts: the Arabian Sea Branch and the Bay of Bengal Branch. The 10 Arabian Sea Branch of the Southwest Monsoon first hits the Western Ghats of the coastal state of Kerala, India, 9 300 thus making the area the first state in India to receive rain from the Southwest Monsoon. This branch of the 8 monsoon moves northwards along the Western Ghats (Konkan and Goa) with precipitation on coastal areas, 7 400 west of the Western Ghats. The eastern areas of the Western Ghats do not receive much rain from this 6 monsoon as the wind does not cross the Western Ghats. 5 500 4 600 3 Southwest Monsoon flows over the Bay of Bengal heading towards North-East India and Bengal, picking up 2 750 more moisture from the Bay of Bengal. The winds arrive at the Eastern Himalayas with large amounts of rain. 1 850 Mawsynram, situated on the southern slopes of the Khasi Hills in Meghalaya, India, is one of the wettest 0 places on Earth. After the arrival at the Eastern Himalayas, the winds turns towards the west, travelling over the Indo-Gangetic Plain at a rate of roughly 1–2 weeks per state, pouring rain all along its way. June 1 is regarded as the date of onset of the monsoon in India, as indicated by the arrival of the monsoon in the The COMET Program southernmost state of Kerala. Figure 2.4: Convective storm (MetEd, 2012) Like flash floods, monsoon floods are often caused by large amounts of rain falling in a short period of time. Flash floods, however, are generally localized. Monsoon rains tend to affect larger areas, causing more damage During this process, moisture is condensing in the rising updraft air and forms rain-drops. The weight of the and higher loss of life. There may be multiple waves of flood during a single monsoon season. Each rainfall raindrops increases and eventually becomes too heavy for the updraft to hold aloft. Subsequently, precipitation (new wave) poses more potential for flooding as large amounts of rain already saturated the ground and rivers will begin to fall back down through the updraft. The resulting rain is of very short duration, seldom more than are overflowing their banks. 1 hour, but the intensities are very high and may amount to 70 to 100 mm. This con-vective type of rain can produce flash floods on streams with small catchments.

Monsoon Weather summary (MetEd 2012):

Weather is complicated, develops across a range of spatial scales, and is ever- In India floods occur mainly during the monsoon as over 80% of India's annual rainfall occur during this changing. Keeping up to date with potential weather hazards requires three things: period. The south-western summer monsoons occur from June through September. The Thar Desert and adjoining areas of the northern and central Indian subcontinent heats up considerably during the hot 1. Knowing what is going on now summers, which causes a low pressure area over the northern and central Indian subcontinent. To fill this 2. Knowing what is expected over the next several days void, the moisture-laden winds from the Indian Ocean rush in to the subcontinent. These winds, rich in 3. Keeping track of what is happening upstream from your area moisture, are drawn towards the Himalayas, creating winds blowing storm clouds towards the subcontinent. The Himalayas act like a high wall, blocking the winds from passing into Central Asia, thus forcing them to rise. With the gain in altitude of the clouds, the temperature drops and precipitation occurs. Some areas of the subcontinent receive up to 10,000 mm (390 inches) of rain annually.

7 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-A: Floods - Hydrologic Basics 8 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best The southwest monsoon is generally expected to begin around the start of June and fade down by the end of

12 km 200 mb September. The moisture-laden winds on reaching the southernmost point of the Indian Peninsula, due to its 11 topography, become divided into two parts: the Arabian Sea Branch and the Bay of Bengal Branch. The 10 Arabian Sea Branch of the Southwest Monsoon first hits the Western Ghats of the coastal state of Kerala, India, 9 300 thus making the area the first state in India to receive rain from the Southwest Monsoon. This branch of the 8 monsoon moves northwards along the Western Ghats (Konkan and Goa) with precipitation on coastal areas, 7 400 west of the Western Ghats. The eastern areas of the Western Ghats do not receive much rain from this 6 monsoon as the wind does not cross the Western Ghats. 5 500 4 600 3 Southwest Monsoon flows over the Bay of Bengal heading towards North-East India and Bengal, picking up 2 750 more moisture from the Bay of Bengal. The winds arrive at the Eastern Himalayas with large amounts of rain. 1 850 Mawsynram, situated on the southern slopes of the Khasi Hills in Meghalaya, India, is one of the wettest 0 places on Earth. After the arrival at the Eastern Himalayas, the winds turns towards the west, travelling over the Indo-Gangetic Plain at a rate of roughly 1–2 weeks per state, pouring rain all along its way. June 1 is regarded as the date of onset of the monsoon in India, as indicated by the arrival of the monsoon in the The COMET Program southernmost state of Kerala. Figure 2.4: Convective storm (MetEd, 2012) Like flash floods, monsoon floods are often caused by large amounts of rain falling in a short period of time. Flash floods, however, are generally localized. Monsoon rains tend to affect larger areas, causing more damage During this process, moisture is condensing in the rising updraft air and forms rain-drops. The weight of the and higher loss of life. There may be multiple waves of flood during a single monsoon season. Each rainfall raindrops increases and eventually becomes too heavy for the updraft to hold aloft. Subsequently, precipitation (new wave) poses more potential for flooding as large amounts of rain already saturated the ground and rivers will begin to fall back down through the updraft. The resulting rain is of very short duration, seldom more than are overflowing their banks. 1 hour, but the intensities are very high and may amount to 70 to 100 mm. This con-vective type of rain can produce flash floods on streams with small catchments.

Monsoon Weather summary (MetEd 2012):

Weather is complicated, develops across a range of spatial scales, and is ever- In India floods occur mainly during the monsoon as over 80% of India's annual rainfall occur during this changing. Keeping up to date with potential weather hazards requires three things: period. The south-western summer monsoons occur from June through September. The Thar Desert and adjoining areas of the northern and central Indian subcontinent heats up considerably during the hot 1. Knowing what is going on now summers, which causes a low pressure area over the northern and central Indian subcontinent. To fill this 2. Knowing what is expected over the next several days void, the moisture-laden winds from the Indian Ocean rush in to the subcontinent. These winds, rich in 3. Keeping track of what is happening upstream from your area moisture, are drawn towards the Himalayas, creating winds blowing storm clouds towards the subcontinent. The Himalayas act like a high wall, blocking the winds from passing into Central Asia, thus forcing them to rise. With the gain in altitude of the clouds, the temperature drops and precipitation occurs. Some areas of the subcontinent receive up to 10,000 mm (390 inches) of rain annually.

7 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-A: Floods - Hydrologic Basics 8 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best 2.2 Influences of land surface 2.4 Snow melt

The properties of the land surface and soil have a major influence on the rainfall loss (this is the part of rain Beside rain, snow melt in winter and in the northern part of India provides a major contribution to which does not contribute to direct runoff). Rainfall losses that are significant in analysing and modelling flood floods. The amount of water stored in the snow pack depends on the condition of the snow pack. One events include land-cover interception, depression storage, and . Evapotranspiration has a negligible centimetre powder snow corresponds to a rainfall of one millimetre of rain, which is one litre of water per effect in most flood events. square meter. Re-peated melting and freezing change the crystal structure of the snow and the density grows. An old snow pack is equivalent up to 4 mm rainfall per cm depth. Interception and depression storage result from the retention of water on the surfaces of vegetation and in local The energy budget of snow pack and its surroundings consists mainly of the following: incoming and reflected depressions in the ground surface. Infiltration (the movement of water into the ground)is the most significant solar radiation, incident, and emitted long wave radiation, turbulent transfer of latent and sensible heat, ground loss during flood events. The infiltration rate is highest among old forest stands and may reach 60-75 mm/hr. conduction, and heat that is advected during rainfall. Under a meadows and pastures it may be 20 mm/h. The soil absorbs water like a sponge. Size and shape of the cavities in the soil determine its storage capacity and the infiltration rate. These vary depending on the The snow albedo or ratio of reflected solar radiation to incident solar radiation, decreases as snow ages, humus content, soil type, soil thickness, and causing the snow pack to absorb more solar energy over time. Latent heat is the energy that is exchanged with density. the surrounding environment during a phase change of water. The latent heat of sublimation is about 8 times greater than the latent heat of fusion. The atmospheric vapour pressure above the snow pack can greatly affect Surface runoff originates either if the rain intensity how much sublimation occurs. Significant warming of the snow pack via and subsequent latent exceeds the infiltration rate or if the total rain heat release usually requires strong winds that induce turbulent transfer. When winds are calm, turbulent amount exceeds the storage capacity of the soil. transfer is at a minimum, and any air that is cooled through sublimation near the snow pack surface remains near the surface. When winds are strong, turbulent transfer occurs, and mixes any cooled air near the snow pack surface with warmer air aloft. This can cause the melting process to begin, especially if winds are strong 2.3 Surface runoff and runoff and vapour pressure increases with height in the atmosphere. concentration B Rain falling on snow pack will impart some heat upon it. The amount of heat imparted depends on the temperature of the rain and any phase change that occurs. A rainfall 'warm' enough that it does not freeze as it Surface runoff is collected by small creeks and B runoff trickles through the snow pack will impart small amounts of heat as it passes through the pack. A rainfall 'cool' rivers. This process is called runoff concentration. enough that it freezes as it trickles through the pack will cause significantly more heat to be gained by the The catchment defines the area from where the snow pack, as latent heat will be released to the snow during freezing. The amount of warming that occurs via water flows to a certain point in the river. The time rainfall freezing within the snow pack depends on the initial density of the snow pack. A of concentration is the time the water needs to A travel from the most distant point in the catchment time When soil is unfrozen and unsaturated, melt water will be absorbed as long as the snow melt rate is less than to the outlet. It depends mainly on the size and the the infiltration rate. When soil is frozen, infiltration of melt water is impeded. This can cause ponding and slope of the catchment. The shape of the possible refreezing of melt water onto the ground. In rapid snow melt situations, this may cause flooding. Figure 2.5: Basins of circular shape respond by catchment has influence on the shape of the higher peak and shorter duration to rainfall In the higher mountains, a thick snow pack is built up over the winter. Nearly all of the precipitation is stored resulting flood . A circular basin than elongated basins (Bayerisches Landesamt as snow. Only a little water is discharged through the rivers. At the onset of warm weather in spring and early produces shorter and higher floods than a basin of fur Wasserwirtschaft, 2004) summer the snow melts off gradually. While in the the zero-degree mark exceeded already, a hundred elongated size.

9 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-A: Floods - Hydrologic Basics 10 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best 2.2 Influences of land surface 2.4 Snow melt

The properties of the land surface and soil have a major influence on the rainfall loss (this is the part of rain Beside rain, snow melt in winter and spring in the northern part of India provides a major contribution to which does not contribute to direct runoff). Rainfall losses that are significant in analysing and modelling flood floods. The amount of water stored in the snow pack depends on the condition of the snow pack. One events include land-cover interception, depression storage, and infiltration. Evapotranspiration has a negligible centimetre powder snow corresponds to a rainfall of one millimetre of rain, which is one litre of water per effect in most flood events. square meter. Re-peated melting and freezing change the crystal structure of the snow and the density grows. An old snow pack is equivalent up to 4 mm rainfall per cm depth. Interception and depression storage result from the retention of water on the surfaces of vegetation and in local The energy budget of snow pack and its surroundings consists mainly of the following: incoming and reflected depressions in the ground surface. Infiltration (the movement of water into the ground)is the most significant solar radiation, incident, and emitted long wave radiation, turbulent transfer of latent and sensible heat, ground loss during flood events. The infiltration rate is highest among old forest stands and may reach 60-75 mm/hr. conduction, and heat that is advected during rainfall. Under a meadows and pastures it may be 20 mm/h. The soil absorbs water like a sponge. Size and shape of the cavities in the soil determine its storage capacity and the infiltration rate. These vary depending on the The snow albedo or ratio of reflected solar radiation to incident solar radiation, decreases as snow ages, humus content, soil type, soil thickness, and causing the snow pack to absorb more solar energy over time. Latent heat is the energy that is exchanged with density. the surrounding environment during a phase change of water. The latent heat of sublimation is about 8 times greater than the latent heat of fusion. The atmospheric vapour pressure above the snow pack can greatly affect Surface runoff originates either if the rain intensity how much sublimation occurs. Significant warming of the snow pack via deposition and subsequent latent exceeds the infiltration rate or if the total rain heat release usually requires strong winds that induce turbulent transfer. When winds are calm, turbulent amount exceeds the storage capacity of the soil. transfer is at a minimum, and any air that is cooled through sublimation near the snow pack surface remains near the surface. When winds are strong, turbulent transfer occurs, and mixes any cooled air near the snow pack surface with warmer air aloft. This can cause the melting process to begin, especially if winds are strong 2.3 Surface runoff and runoff and vapour pressure increases with height in the atmosphere. concentration B Rain falling on snow pack will impart some heat upon it. The amount of heat imparted depends on the temperature of the rain and any phase change that occurs. A rainfall 'warm' enough that it does not freeze as it Surface runoff is collected by small creeks and B runoff trickles through the snow pack will impart small amounts of heat as it passes through the pack. A rainfall 'cool' rivers. This process is called runoff concentration. enough that it freezes as it trickles through the pack will cause significantly more heat to be gained by the The catchment defines the area from where the snow pack, as latent heat will be released to the snow during freezing. The amount of warming that occurs via water flows to a certain point in the river. The time rainfall freezing within the snow pack depends on the initial density of the snow pack. A of concentration is the time the water needs to A travel from the most distant point in the catchment time When soil is unfrozen and unsaturated, melt water will be absorbed as long as the snow melt rate is less than to the outlet. It depends mainly on the size and the the infiltration rate. When soil is frozen, infiltration of melt water is impeded. This can cause ponding and slope of the catchment. The shape of the possible refreezing of melt water onto the ground. In rapid snow melt situations, this may cause flooding. Figure 2.5: Basins of circular shape respond by catchment has influence on the shape of the higher peak and shorter duration to rainfall In the higher mountains, a thick snow pack is built up over the winter. Nearly all of the precipitation is stored resulting flood hydrograph. A circular basin than elongated basins (Bayerisches Landesamt as snow. Only a little water is discharged through the rivers. At the onset of warm weather in spring and early produces shorter and higher floods than a basin of fur Wasserwirtschaft, 2004) summer the snow melts off gradually. While in the valley the zero-degree mark exceeded already, a hundred elongated size.

9 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-A: Floods - Hydrologic Basics 10 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best yards further up the slope there is still the frost. Only slowly the melting proceeds to greater heights. In the best

case, the snow has melted in the lower regions and the water is already drained before the snow melts in the 3 3000 Discharge in m /s higher altitudes.

2500 In the plains and hills no snow pack capable of storage is built up. The frequent changes between snowfall and rain has compacted the snow to high density. Because of the lower height differences, larger areas can be 2000 affected by snow melt at the same time releasing large amounts of water. Melting in the low mountain ranges is usually initiated by moist and mild air masses. 1500

2.5 Channel flow and flood routing 1000

300 Flood routing is the process of movement of a flood wave through a river system. The effects of storage and flow resistance are reflected in the shape and timing of of the wave at different locations 0 10. 11. 12. 14.13. 15. 16. 1817. 19. 20. 21. 22. 23.23. 24. 25. 26. 27. 28. 29. 30. 31. 1. downstream. The travel time of the flood peak downstream is used for flood warning since ancient times. Mai 1999 Juni in have a major influence on the travel time and height of the flood peak. If two rivers Figure 2.7: Development of a flood wave downstream by flood routing and tributaries. The example shows the flood of May, 1999 in the Bavarian part of the Danube River. Between the gauges of Ingolstadt and when both are at the flood stage, the combination of flows can lead to a jump from a medium to a Kelheim there was a dike breach clearly visible by the saddle near the peak in the hydrograph of Kelheim high flood. (Bayerisches Landesamt für Wasserwirtschaft, 2004).

Water level

Flood peak Important points to remember about flood science

Hydrograph ¢ at gauge site A A combination of high rainfall rate and very efficient runoff production is common to most flash flood Hydrograph events. at gauge site B ¢ In some situations the runoff characteristics can be as or more important than the rain rate. ¢ Soil moisture, soil permeability, soil surface alterations, and vertical soil profile are important soil characteristics that affect runoff production and hence help define flash flood prone areas. ¢ Basin characteristics, (e.g., size, shape, slope, land cover) influence runoff and hence flash flood occurrence potential. Travel time of peak ¢ Gauge site A Urbanization and fire can greatly increase flash flood potential by increasing both the potential volume of runoff as well as the speed with which the runoff occurs. Gauge site B (National Weather Service, 2010)

Questions

Figure 2.6: Flood wave moving downstream without lateral inflow. The peak is damped by retention in How much rain leads to flooding where? the river bed and in the flood plain. The travel time is related to the mean velocity of the water depending on slope and roughness (Bayerisches Landesamt für Wasserwirtschaft, 2004). How do basin and soil characteristics impact the generation of surface runoff? How will the water move once it reaches the ground?

11 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-A: Floods - Hydrologic Basics 12 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best yards further up the slope there is still the frost. Only slowly the melting proceeds to greater heights. In the best case, the snow has melted in the lower regions and the water is already drained before the snow melts in the 3 3000 Discharge in m /s higher altitudes.

2500 In the plains and hills no snow pack capable of storage is built up. The frequent changes between snowfall and rain has compacted the snow to high density. Because of the lower height differences, larger areas can be 2000 affected by snow melt at the same time releasing large amounts of water. Melting in the low mountain ranges is usually initiated by moist and mild air masses. 1500

2.5 Channel flow and flood routing 1000

300 Flood routing is the process of movement of a flood wave through a river system. The effects of storage and flow resistance are reflected in the shape and timing of hydrographs of the wave at different locations 0 10. 11. 12. 14.13. 15. 16. 1817. 19. 20. 21. 22. 23.23. 24. 25. 26. 27. 28. 29. 30. 31. 1. downstream. The travel time of the flood peak downstream is used for flood warning since ancient times. Mai 1999 Juni Tributaries in flood stage have a major influence on the travel time and height of the flood peak. If two rivers Figure 2.7: Development of a flood wave downstream by flood routing and tributaries. The example shows the flood of May, 1999 in the Bavarian part of the Danube River. Between the gauges of Ingolstadt and confluence when both are at the flood stage, the combination of flows can lead to a jump from a medium to a Kelheim there was a dike breach clearly visible by the saddle near the peak in the hydrograph of Kelheim high flood. (Bayerisches Landesamt für Wasserwirtschaft, 2004).

Water level

Flood peak Important points to remember about flood science

Hydrograph ¢ at gauge site A A combination of high rainfall rate and very efficient runoff production is common to most flash flood Hydrograph events. at gauge site B ¢ In some situations the runoff characteristics can be as or more important than the rain rate. ¢ Soil moisture, soil permeability, soil surface alterations, and vertical soil profile are important soil characteristics that affect runoff production and hence help define flash flood prone areas. ¢ Basin characteristics, (e.g., size, shape, slope, land cover) influence runoff and hence flash flood occurrence potential. Travel time of peak ¢ Gauge site A Urbanization and fire can greatly increase flash flood potential by increasing both the potential volume of runoff as well as the speed with which the runoff occurs. Gauge site B (National Weather Service, 2010)

Questions

Figure 2.6: Flood wave moving downstream without lateral inflow. The peak is damped by retention in How much rain leads to flooding where? the river bed and in the flood plain. The travel time is related to the mean velocity of the water depending on slope and roughness (Bayerisches Landesamt für Wasserwirtschaft, 2004). How do basin and soil characteristics impact the generation of surface runoff? How will the water move once it reaches the ground?

11 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-A: Floods - Hydrologic Basics 12 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best The primary goal of a warning system is to prevent hazards from ¢ detection, becoming disasters National Weather Service (2010). ¢ warning, and

There is a high diversity of flood warning systems in operation. Flood ¢ response warning systems are typically tailor-made to suit specific requirements for the location for which the warnings are to be In the detection stage, real time data on processes that could generate a flood event are monitored. This provided, ranging from fast-responding local warning systems in the includes primarily, monitoring of hydrological and meteorological conditions in the catchment through on-line headwaters of a river or urban areas to flood warning systems for information gathered through telemetry systems, climate stations, weather radar, and so on. In addition the lower reaches of large river basins. For example manual self-help detection stage may include the forecast system to increase the lead time of the warnings. systems (comprised of a local data collection system, community 3 In the forecasting stage predictions are made of levels and flows, as well as the time of occurrence of possible flood coordinator, flood forecast procedure, communication network to forthcoming flood events. Typically, this involves the use of hydrological models, driven by both the real-time distribute warnings and a response plan) are inexpensive and simple LU-B: data gathered in the detection phase and forecasts of meteorological conditions such as rainfall and to operate but may not have the best temporal resolution needed for temperature. These are often obtained through external meteorological forecasts. Forecasting may also include short-duration accumulation and rainfall rates. The most simple flash now casting techniques such as storm tracking or propagation of moving rain fields in weather radar Flood flood alarm system consists of a water-level sensor(s) connected to an observations (Werner et al., 2006). audible and/or visible alarm device located at a community agency with 24-hour operation. Warning Flood forecast and warning may be triggered by monitoring rain and snow melt, water level or forecast of rain and snow melt. Threshold values have to be derived from the conditions of observed flood stages. Systems Flood Warning Systems can be operated on a local, regional, basin wide, national, or even continental base like the Flash Flood The warning stage is a key factor in the success of operational flood warning. Using information derived from Guidance (WMO, 2007). Often, flood warning systems are developed the detection and forecasting stages, the decision to warn appropriate authorities and/or properties at risk must to cover all rivers within an administrative boundary, depending on be taken. The warning must be such that it gives an unambiguous message on the imminent flood potential. the responsibilities of the authorities whose obligation it is to deliver the warnings. Response to flood warnings issued is vital for achieving the aims of operational flood warning. If the objective is to reduce damage through flood preparedness, an appropriate response by relevant authorities and affected The organization and structure of the system in adjacent regions or persons must be taken following a warning. countries may be very different. In some cases, differences may even occur within the same river system, such as in the Rhine basin. No Although flood forecasting is an important part of the flood warning process, the most basic warning systems less than 25 flood forecasting and/or flood warning centres are do not include an explicit forecasting step even nowadays. The basic systems issue flood warnings on the involved from the source to mouth of the River Rhine, each with a basis of observations such as gauged rainfall and flows, combined with the judgment and experience of the different approach to flood forecasting. forecasters.

Despite the apparent variety in approaches, the basic elements on Installing this basic system has to be the prior step before implementation of the forecast system. The goals of which the effective flood warning depends are: the forecast cannot be reached if the basic flood warning system is not implemented in a proper working manner.

13 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-B: Flood Warning Systems 14 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best The primary goal of a warning system is to prevent hazards from ¢ detection, becoming disasters National Weather Service (2010). ¢ warning, and

There is a high diversity of flood warning systems in operation. Flood ¢ response warning systems are typically tailor-made to suit specific requirements for the location for which the warnings are to be In the detection stage, real time data on processes that could generate a flood event are monitored. This provided, ranging from fast-responding local warning systems in the includes primarily, monitoring of hydrological and meteorological conditions in the catchment through on-line headwaters of a river or urban areas to flood warning systems for information gathered through telemetry systems, climate stations, weather radar, and so on. In addition the lower reaches of large river basins. For example manual self-help detection stage may include the forecast system to increase the lead time of the warnings. systems (comprised of a local data collection system, community 3 In the forecasting stage predictions are made of levels and flows, as well as the time of occurrence of possible flood coordinator, flood forecast procedure, communication network to forthcoming flood events. Typically, this involves the use of hydrological models, driven by both the real-time distribute warnings and a response plan) are inexpensive and simple LU-B: data gathered in the detection phase and forecasts of meteorological conditions such as rainfall and to operate but may not have the best temporal resolution needed for temperature. These are often obtained through external meteorological forecasts. Forecasting may also include short-duration accumulation and rainfall rates. The most simple flash now casting techniques such as storm tracking or propagation of moving rain fields in weather radar Flood flood alarm system consists of a water-level sensor(s) connected to an observations (Werner et al., 2006). audible and/or visible alarm device located at a community agency with 24-hour operation. Warning Flood forecast and warning may be triggered by monitoring rain and snow melt, water level or forecast of rain and snow melt. Threshold values have to be derived from the conditions of observed flood stages. Systems Flood Warning Systems can be operated on a local, regional, basin wide, national, or even continental base like the Flash Flood The warning stage is a key factor in the success of operational flood warning. Using information derived from Guidance (WMO, 2007). Often, flood warning systems are developed the detection and forecasting stages, the decision to warn appropriate authorities and/or properties at risk must to cover all rivers within an administrative boundary, depending on be taken. The warning must be such that it gives an unambiguous message on the imminent flood potential. the responsibilities of the authorities whose obligation it is to deliver the warnings. Response to flood warnings issued is vital for achieving the aims of operational flood warning. If the objective is to reduce damage through flood preparedness, an appropriate response by relevant authorities and affected The organization and structure of the system in adjacent regions or persons must be taken following a warning. countries may be very different. In some cases, differences may even occur within the same river system, such as in the Rhine basin. No Although flood forecasting is an important part of the flood warning process, the most basic warning systems less than 25 flood forecasting and/or flood warning centres are do not include an explicit forecasting step even nowadays. The basic systems issue flood warnings on the involved from the source to mouth of the River Rhine, each with a basis of observations such as gauged rainfall and flows, combined with the judgment and experience of the different approach to flood forecasting. forecasters.

Despite the apparent variety in approaches, the basic elements on Installing this basic system has to be the prior step before implementation of the forecast system. The goals of which the effective flood warning depends are: the forecast cannot be reached if the basic flood warning system is not implemented in a proper working manner.

13 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-B: Flood Warning Systems 14 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best For this case, the local authorities have plans of endangered areas or buildings and plans for the organisation Real-Time Data On-Line Data Automated Product / Data Clients Collection Base Products Dissemination of flood defence systems. Public Polled Systems Metadata Forecasts Warnings Emergency (Actively Sampling) Waterlevel Maps WWW Local Gov NFS File Tranfer & Discharge Bulletins FPT Water Auth. The flood warnings have to be actively transmitted to the concerned recipients. Once warned, the recipients FTP & http-request Meteodata E-Mails, SMS Others have the responsibility to inform themselves about the threatening flood. For this purpose, the state offices for water management, the regional flood forecast centres and the flood information centre provide updated data, forecasts and flood reports to the public and authorities via the website as well as flood news via a telephone Hydromet Input Analysis & service. Actual water levels of all gauging stations are provided via TV-text and telephone service. Modeling NWP Snowforecast Data Viewing Radar Hydrological Modeling Meteorologists Flood Risk Assesment Flood Warning Service

Figure 3.1: Components of a flood warning system (after Elliot et al., 2005) Flood Forecasts 5 Regional Flood Warning Centre / The Figure 3.1 shows the scheme of an on-line operational system used for flood warning. The figure follows Forecast Bavarian Environment Centres Agency Status report for Bavaria Elliot et al. (2005) of the Australian Bureau of Meteorology with some minor modifications, but it can be taken Website of Floodline, the Flood as an example for other flood warning services, too. The upper series is a full automated chain beginning by Teletext Warning real-time data collection for detection, sampling the data in a central data base, creating products like maps, State offices for Service water management Water Levels forecasts, warnings and disseminate the products to the clients for warning. Regional Warnings 3.1 Bavarian flood information service Country district offices

Bavaria has set up a Flood Warning Service that follows the scheme in Figure 3.1 quite closely. It collects data Towns & Communities Private Citizens of water levels, runoff and precipitation, analyses this information, draws up flood alert plans and warns people affected. Connected to the Flood Warning Service are the state offices for water management, the Internal communications Alert channels Available information county district offices, towns, and communities. The coordinating unit is the Flood Warning Centre in the Bavarian Environment Agency (Figure 3.2). Five regional Flood Forecast Centres are responsible for calculating Figure 3.2: Reporting and information scheme of the Bavarian flood and preparing flood forecasts for the basins of the rivers Main, Danube, Isar, Iller/Lech and Inn respectively. information service (Bayerisches Landesamt für Umwelt, 2011)

The respective centres are put into action as soon as rivers or lakes rise above defined threshold values. The Important points to remember about Flood Early Warning Systems water levels at the gauging stations are then read on an hourly basis and flood predictions are updated continuously. The state offices for water management put out regional alerts, while the Flood Warning Centre ¢ One of the first steps in designing an Early Warning System (EWS) is to assess existing capability (and issues a flood status report for all Bavaria. infrastructure) that could be employed by the EWS and what gaps need to be filled by new capability and infrastructure. Alert plans make sure that this information is passed on through the county district offices to the affected ¢ Lack of forecast skill makes it very difficult to issue flash flood warnings several hours in advance based towns and communities. Towns and communities are the last link in the alert channel and play a particularly upon Quantitative Precipitation Forecasts (QPF). However, flash flood watches can be issued based upon QPFs and then later modified as necessary based upon observed rainfall (QPE). important role. The alert plans specify who is to be warned, when and how, and what measures are to be (National Weather Service, 2010) taken at which gauge levels.

15 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-B: Flood Warning Systems 16 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best For this case, the local authorities have plans of endangered areas or buildings and plans for the organisation Real-Time Data On-Line Data Automated Product / Data Clients Collection Base Products Dissemination of flood defence systems. Public Polled Systems Metadata Forecasts Warnings Emergency (Actively Sampling) Waterlevel Maps WWW Local Gov NFS File Tranfer & Discharge Bulletins FPT Water Auth. The flood warnings have to be actively transmitted to the concerned recipients. Once warned, the recipients FTP & http-request Meteodata E-Mails, SMS Others have the responsibility to inform themselves about the threatening flood. For this purpose, the state offices for water management, the regional flood forecast centres and the flood information centre provide updated data, forecasts and flood reports to the public and authorities via the website as well as flood news via a telephone Hydromet Input Analysis & service. Actual water levels of all gauging stations are provided via TV-text and telephone service. Modeling NWP Snowforecast Data Viewing Radar Hydrological Modeling Meteorologists Flood Risk Assesment Flood Warning Service

Figure 3.1: Components of a flood warning system (after Elliot et al., 2005) Flood Forecasts 5 Regional Flood Warning Centre / The Figure 3.1 shows the scheme of an on-line operational system used for flood warning. The figure follows Forecast Bavarian Environment Centres Agency Status report for Bavaria Elliot et al. (2005) of the Australian Bureau of Meteorology with some minor modifications, but it can be taken Website of Floodline, the Flood as an example for other flood warning services, too. The upper series is a full automated chain beginning by Teletext Warning real-time data collection for detection, sampling the data in a central data base, creating products like maps, State offices for Service water management Water Levels forecasts, warnings and disseminate the products to the clients for warning. Regional Warnings 3.1 Bavarian flood information service Country district offices

Bavaria has set up a Flood Warning Service that follows the scheme in Figure 3.1 quite closely. It collects data Towns & Communities Private Citizens of water levels, runoff and precipitation, analyses this information, draws up flood alert plans and warns people affected. Connected to the Flood Warning Service are the state offices for water management, the Internal communications Alert channels Available information county district offices, towns, and communities. The coordinating unit is the Flood Warning Centre in the Bavarian Environment Agency (Figure 3.2). Five regional Flood Forecast Centres are responsible for calculating Figure 3.2: Reporting and information scheme of the Bavarian flood and preparing flood forecasts for the basins of the rivers Main, Danube, Isar, Iller/Lech and Inn respectively. information service (Bayerisches Landesamt für Umwelt, 2011)

The respective centres are put into action as soon as rivers or lakes rise above defined threshold values. The Important points to remember about Flood Early Warning Systems water levels at the gauging stations are then read on an hourly basis and flood predictions are updated continuously. The state offices for water management put out regional alerts, while the Flood Warning Centre ¢ One of the first steps in designing an Early Warning System (EWS) is to assess existing capability (and issues a flood status report for all Bavaria. infrastructure) that could be employed by the EWS and what gaps need to be filled by new capability and infrastructure. Alert plans make sure that this information is passed on through the county district offices to the affected ¢ Lack of forecast skill makes it very difficult to issue flash flood warnings several hours in advance based towns and communities. Towns and communities are the last link in the alert channel and play a particularly upon Quantitative Precipitation Forecasts (QPF). However, flash flood watches can be issued based upon QPFs and then later modified as necessary based upon observed rainfall (QPE). important role. The alert plans specify who is to be warned, when and how, and what measures are to be (National Weather Service, 2010) taken at which gauge levels.

15 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-B: Flood Warning Systems 16 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best 3.2 Water level and flood risk in Bavaria Table 3.1: Flood hazards and protective measures in relation to the water level at gauging Station Passau Danube (cut version) (website www.hnd.bayern.de)

The basic information in flood warning is the water level at the gauge site measured or forecasted. For W cm Place Type of measure or hazard interpretation in respect to the flood risk and to protective measures more information is needed. First, there are 4 alert levels that deliver plain and simple information on the extent of flooding. At each individual gauging 630 Passau Advance warning to the police by the regulatory agency station it is determined which water levels correspond to the respective flood alert levels (Figure 3.3). Second, 720 Passau Flooding of the upper Zollände in Passau and the harbour area Racklau. there are tables for each gauging station connecting water level to a description of the flood extent, protective Flooding of the Fritz-Schäffer-Promenade at the tavern -Tiroler-, removal of the 720 Passau measures and flood hazard (Table 3.1).The communities are responsible for the maintenance of plans linking motor vehicles. water level to flood risk and protective response. They keep plans of endangered areas or buildings and plans 740 Passau Flooding of the Fritz-Schäffer-Promenade, traffic stoppage. for the organisation of flood defence systems. Protective responses in Bavaria are: manageable flood polders, major storage dams in headwater, flood protection walls, mobile closures, superstructures, sandbags and 750 Passau Evacuation of the parking lots at Schanzl. evacuation. 750 Passau Traffic blocking at the upper Donaulände

770 Passau Flooding of the high road ST 2132 at Löwmühle Evacuation of garages and souvenir shops at the lower Donaulände. flooding of the 770 Passau harbour area Racklau 1 2 3 4 780 Passau Highest water lever for water navigation (HSW).

790 Passau Flooding of the approach to the Nagelschmied- and Höllgasse

800 Passau Flooding of the posterior Donaulände downstream of the Schanzl bridge. Warning by loudspeaker for the upper Donaulände, Regensburger street, 800 Passau Rindermarkt and Ort. 810 Passau ‘Ort’ partly flooded.

3.3 Website of the Bavarian Flood Information Service Alert levels indicate Alert level 1: Minor overflows in some areas the specific flood Alert level 2: Flooding of farmland/undeveloped The website of the Flood Warning Service (Figure 3.4) gives access to detailed background information and to status areas or minor hindrance to traffic on main roads and local roads the latest water level readings and recorded measurements. Maps and charts give a quick overview of the Alert level 3: Flooding of some buildings or cellars status 24 hours a day. These sites give access to water level and run-off data from river and lake or impassable inter-regional roads or flood mitigation monitoring stations as well as to measurements provided by precipitation and snow depth monitoring stations. task forces required in some areas In a flood event the Flood Bulletin (status report) gives an overview of the flood situation and a forecast of the Alert level 4: Extensive flooding of built-up areas or flood mitigation task forces required on a large scale expected further development. The bulletin is updated several times a day. In their Warnings the state offices for water management provide a detailed description of current and predicted flooding for each one of their Figure 3.3: Floodalert levels (Bayerisches Landesamt für Umwelt, 2011) county districts that are threatened. In addition reports on past flood events as well as links, addresses and phone numbers of other contact partners can be found on the website (www.hnd.bayern.de).

17 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-B: Flood Warning Systems 18 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best 3.2 Water level and flood risk in Bavaria Table 3.1: Flood hazards and protective measures in relation to the water level at gauging Station Passau Danube (cut version) (website www.hnd.bayern.de)

The basic information in flood warning is the water level at the gauge site measured or forecasted. For W cm Place Type of measure or hazard interpretation in respect to the flood risk and to protective measures more information is needed. First, there are 4 alert levels that deliver plain and simple information on the extent of flooding. At each individual gauging 630 Passau Advance warning to the police by the regulatory agency station it is determined which water levels correspond to the respective flood alert levels (Figure 3.3). Second, 720 Passau Flooding of the upper Zollände in Passau and the harbour area Racklau. there are tables for each gauging station connecting water level to a description of the flood extent, protective Flooding of the Fritz-Schäffer-Promenade at the tavern -Tiroler-, removal of the 720 Passau measures and flood hazard (Table 3.1).The communities are responsible for the maintenance of plans linking motor vehicles. water level to flood risk and protective response. They keep plans of endangered areas or buildings and plans 740 Passau Flooding of the Fritz-Schäffer-Promenade, traffic stoppage. for the organisation of flood defence systems. Protective responses in Bavaria are: manageable flood polders, major storage dams in headwater, flood protection walls, mobile closures, superstructures, sandbags and 750 Passau Evacuation of the parking lots at Schanzl. evacuation. 750 Passau Traffic blocking at the upper Donaulände

770 Passau Flooding of the high road ST 2132 at Löwmühle Evacuation of garages and souvenir shops at the lower Donaulände. flooding of the 770 Passau harbour area Racklau 1 2 3 4 780 Passau Highest water lever for water navigation (HSW).

790 Passau Flooding of the approach to the Nagelschmied- and Höllgasse

800 Passau Flooding of the posterior Donaulände downstream of the Schanzl bridge. Warning by loudspeaker for the upper Donaulände, Regensburger street, 800 Passau Rindermarkt and Ort. 810 Passau ‘Ort’ partly flooded.

3.3 Website of the Bavarian Flood Information Service Alert levels indicate Alert level 1: Minor overflows in some areas the specific flood Alert level 2: Flooding of farmland/undeveloped The website of the Flood Warning Service (Figure 3.4) gives access to detailed background information and to status areas or minor hindrance to traffic on main roads and local roads the latest water level readings and recorded measurements. Maps and charts give a quick overview of the Alert level 3: Flooding of some buildings or cellars current status 24 hours a day. These sites give access to water level and run-off data from river and lake or impassable inter-regional roads or flood mitigation monitoring stations as well as to measurements provided by precipitation and snow depth monitoring stations. task forces required in some areas In a flood event the Flood Bulletin (status report) gives an overview of the flood situation and a forecast of the Alert level 4: Extensive flooding of built-up areas or flood mitigation task forces required on a large scale expected further development. The bulletin is updated several times a day. In their Warnings the state offices for water management provide a detailed description of current and predicted flooding for each one of their Figure 3.3: Floodalert levels (Bayerisches Landesamt für Umwelt, 2011) county districts that are threatened. In addition reports on past flood events as well as links, addresses and phone numbers of other contact partners can be found on the website (www.hnd.bayern.de).

17 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-B: Flood Warning Systems 18 Forecasts Hydrological Requirements Requirements and Hydrological Forecasts

Hydrological Best Hydrological Forecasts Best Practice Practice Forecasts and Hydrological Best Hydrological Best

Figure 3.4: Website of the Bavarian Flood Information Service

Questions

What are the major components of a flood warning system?

What is the goal of a flood warning system?

What are the benefits and costs of flood warning system?

19 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice Forecasts Hydrological Requirements Requirements and Hydrological Forecasts

Hydrological Best Hydrological Forecasts Best Practice Practice Forecasts and Hydrological Best Hydrological Best

Figure 3.4: Website of the Bavarian Flood Information Service

Questions

What are the major components of a flood warning system?

What is the goal of a flood warning system?

What are the benefits and costs of flood warning system?

19 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best Monitoring of hydro-meteorological conditions in the catchment is of simply be through telephone or where reliability of telephone networks is insufficient, through dedicated VHS topmost importance for the operational flood warning and forecasting or UHF radio links. The advances in communication technology have in recent years significantly increased the system. The availability of real-time observed data is also important reliability of data transfer, with off-the-shelf technology such as GSM networks and Internet providing reliable during the flood forecasting stage, as these provide the boundary transfer mechanisms (Werner et al. 2006). conditions to the models used. The observed data is used not only to bring the models up-to-date, but also for keeping the models in track with reality through data assimilation. Particularly where lag times are such that desired lead times can be achieved through calculating runoff and routing of observed rainfall and flow, the availability of real- Hofkirchen time data is essential (Werner et. al., 2006). Donau 4 (cm) 400 300 200 100 0 LU-C: Sa so Mo D 14.08.12, 12:00 Meldestufe: 0 Wasserstand (cm): 204 Real Time Abfluss (m2 /s): 323,00 Data Acquisition Meldestufen: keine 1 2 3 4 ohne Meldestufen Derzeit Keine Daten x Panel mit Vorhersage

and Figure 4.2: Presentation of gauging station with actual flood stages marked by different colours (http://www.hnd.bayern.de) Figure 4.1: Real time gauging station (flood stage) Monitoring (http://www.hnd.bayern.de) Although the problems of data transmission are being resolved through advances in communication technology, the reliability of the data itself is an issue that warrants close attention. Validation and checking of Network Most monitoring data is typically provided from telemetry sites, all real-time data prior to their use in a forecasting model is important to avoid warnings being issued on the including river gauges and rain gauges. These sites are used for a basis of doubtful data. more general purpose than flood warning alone, providing data also to navigation authorities, water resources planners, irrigation agencies, The data used is mainly that of rainfall accumulations from rain gauges and levels from river gauging stations. and others. For upstream boundaries, water levels are typically transformed into discharge through stage discharge relationships. Particularly in severe floods, these stage-discharge relationships may contain a significant degree For flood warning, the timely availability of the data is obviously an of error, and the reliability of these relationships at high flows should be carefully assessed. For catchment important factor in determining the utility of the data, and also rainfall estimates derived from rain gauges, the problem of scale arises, as the point measurements must be reliability of data transfer is an important aspect. Communication may up-scaled to catchment rainfall estimates (Werner et al. 2006).

21 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-C: Real Time Data Acquisition and Monitoring Network 22 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best Monitoring of hydro-meteorological conditions in the catchment is of simply be through telephone or where reliability of telephone networks is insufficient, through dedicated VHS topmost importance for the operational flood warning and forecasting or UHF radio links. The advances in communication technology have in recent years significantly increased the system. The availability of real-time observed data is also important reliability of data transfer, with off-the-shelf technology such as GSM networks and Internet providing reliable during the flood forecasting stage, as these provide the boundary transfer mechanisms (Werner et al. 2006). conditions to the models used. The observed data is used not only to bring the models up-to-date, but also for keeping the models in track with reality through data assimilation. Particularly where lag times are such that desired lead times can be achieved through calculating runoff and routing of observed rainfall and flow, the availability of real- Hofkirchen time data is essential (Werner et. al., 2006). Donau 4 (cm) 400 300 200 100 0 LU-C: Sa so Mo D 14.08.12, 12:00 Meldestufe: 0 Wasserstand (cm): 204 Real Time Abfluss (m2 /s): 323,00 Data Acquisition Meldestufen: keine 1 2 3 4 ohne Meldestufen Derzeit Keine Daten x Panel mit Vorhersage and Figure 4.2: Presentation of gauging station with actual flood stages marked by different colours (http://www.hnd.bayern.de) Figure 4.1: Real time gauging station (flood stage) Monitoring (http://www.hnd.bayern.de) Although the problems of data transmission are being resolved through advances in communication technology, the reliability of the data itself is an issue that warrants close attention. Validation and checking of Network Most monitoring data is typically provided from telemetry sites, all real-time data prior to their use in a forecasting model is important to avoid warnings being issued on the including river gauges and rain gauges. These sites are used for a basis of doubtful data. more general purpose than flood warning alone, providing data also to navigation authorities, water resources planners, irrigation agencies, The data used is mainly that of rainfall accumulations from rain gauges and levels from river gauging stations. and others. For upstream boundaries, water levels are typically transformed into discharge through stage discharge relationships. Particularly in severe floods, these stage-discharge relationships may contain a significant degree For flood warning, the timely availability of the data is obviously an of error, and the reliability of these relationships at high flows should be carefully assessed. For catchment important factor in determining the utility of the data, and also rainfall estimates derived from rain gauges, the problem of scale arises, as the point measurements must be reliability of data transfer is an important aspect. Communication may up-scaled to catchment rainfall estimates (Werner et al. 2006).

21 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-C: Real Time Data Acquisition and Monitoring Network 22 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best Important points to remember about gauge networks ¢ Robust communication between the observation networks and the forecasting centre are the key to the success of a flood warning system. In adopting a communication system for a gauge installation, one consideration has to be its reliability under severe environmental conditions. ¢ Alternate communication paths for data collection and product dissemination are needed within a Meteorological and Hydrological Service to ensure 24/7 operations. (National Weather Service, 2010)

4.1 Satellite-based estimates of rainfall and soil moisture

The finest resolution of satellite-based precipitation estimates is currently around 4 km, which is about 3 times coarser than even the largest range gates on most radar.

Figure 4.3: Real-time precipitation-station (www.hnd.bayern.de) Another limitation of precipitation rates estimations by geostationary satellite is that infrared cloud surface temperatures are used only as a proxy for a cloud that is producing precipitation and is not a measure of the precipitation itself. Thus, the algorithms sometimes mis diagnose low to medium level clouds because of their higher temperature as not producing precipitation even when they are. Furthermore the falling rain can be The use of weather radar partially solves the problem of spatial coverage of rainfall estimates. Weather radar drifted by winds and hits a different location than assumed by the analysis of the satellite data. For this provides a spatial estimate of precipitation as inferred through the radar reflectivity of the rainfall, and a reliable reasons, the used algorithms are best suited for deep convection situations in the tropics and mid-latitudes, relationship between reflectivity and rainfall rate must be applied. Unfortunately, no unique relationship is especially when the environmental winds do not vary much with height. available. In addition, there are a number of physical factors influencing uncertainties in radar derived rainfall

estimates. These are not only due to atmospheric distortions such as the melting layer, but also due to, for Products that blend microwave precipitation rates from polar orbiting satellites with geostationary cloud-top example, radar beam shielding in mountainous areas. In practice, integration of rain gauge data and radar temperature information are usually more accurate than a simple geostationary estimate because they are rainfall estimates can be combined to provide more reliable estimates of distribution and volume of observed based on direct observation of emissions from precipitation hydrometeors. But, their reduced spatial and rainfall (Werner et. al., 2006). temporal coverage means that true microwave rain rate products for any given location become available only every 3 to 4 hours on average, which sometimes limits is usefulness for detecting the severity and timing of Important points to remember about gauge networks flash flood type situations.

¢ Accurate, reliable, and timely gauge data is essential to the success of a flood warning system. Important points to remember about satellite data ¢ The establishment of a hydro-meteorological network for flood support can often be an exercise in ¢ Satellite estimates of precipitation can be partially corrected by simultaneously measured rain gauge to integrating and automating already existing networks. get “ground truth” data. ¢ Tipping-bucket rain gauges tend to underestimate precipitation during periods of intense rainfall, but ¢ Gridded rainfall estimates are the primary source of precipitation information for areas that lack radar they are less expensive than weighing gauges. networks and networks of rain gauges. ¢ All gauges have a maintenance cost, although the cost associated with tipping bucket gauges is less ¢ Rainfall estimates can be computed for the entire planet by making use of both Polar Orbiting and than that associated with weighing gauges. Geostationary satellites. (National Weather Service, 2010) (National Weather Service, 2010)

23 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-C: Real Time Data Acquisition and Monitoring Network 24 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best Important points to remember about gauge networks ¢ Robust communication between the observation networks and the forecasting centre are the key to the success of a flood warning system. In adopting a communication system for a gauge installation, one consideration has to be its reliability under severe environmental conditions. ¢ Alternate communication paths for data collection and product dissemination are needed within a Meteorological and Hydrological Service to ensure 24/7 operations. (National Weather Service, 2010)

4.1 Satellite-based estimates of rainfall and soil moisture

The finest resolution of satellite-based precipitation estimates is currently around 4 km, which is about 3 times coarser than even the largest range gates on most radar.

Figure 4.3: Real-time precipitation-station (www.hnd.bayern.de) Another limitation of precipitation rates estimations by geostationary satellite is that infrared cloud surface temperatures are used only as a proxy for a cloud that is producing precipitation and is not a measure of the precipitation itself. Thus, the algorithms sometimes mis diagnose low to medium level clouds because of their higher temperature as not producing precipitation even when they are. Furthermore the falling rain can be The use of weather radar partially solves the problem of spatial coverage of rainfall estimates. Weather radar drifted by winds and hits a different location than assumed by the analysis of the satellite data. For this provides a spatial estimate of precipitation as inferred through the radar reflectivity of the rainfall, and a reliable reasons, the used algorithms are best suited for deep convection situations in the tropics and mid-latitudes, relationship between reflectivity and rainfall rate must be applied. Unfortunately, no unique relationship is especially when the environmental winds do not vary much with height. available. In addition, there are a number of physical factors influencing uncertainties in radar derived rainfall estimates. These are not only due to atmospheric distortions such as the melting layer, but also due to, for Products that blend microwave precipitation rates from polar orbiting satellites with geostationary cloud-top example, radar beam shielding in mountainous areas. In practice, integration of rain gauge data and radar temperature information are usually more accurate than a simple geostationary estimate because they are rainfall estimates can be combined to provide more reliable estimates of distribution and volume of observed based on direct observation of emissions from precipitation hydrometeors. But, their reduced spatial and rainfall (Werner et. al., 2006). temporal coverage means that true microwave rain rate products for any given location become available only every 3 to 4 hours on average, which sometimes limits is usefulness for detecting the severity and timing of Important points to remember about gauge networks flash flood type situations.

¢ Accurate, reliable, and timely gauge data is essential to the success of a flood warning system. Important points to remember about satellite data ¢ The establishment of a hydro-meteorological network for flood support can often be an exercise in ¢ Satellite estimates of precipitation can be partially corrected by simultaneously measured rain gauge to integrating and automating already existing networks. get “ground truth” data. ¢ Tipping-bucket rain gauges tend to underestimate precipitation during periods of intense rainfall, but ¢ Gridded rainfall estimates are the primary source of precipitation information for areas that lack radar they are less expensive than weighing gauges. networks and networks of rain gauges. ¢ All gauges have a maintenance cost, although the cost associated with tipping bucket gauges is less ¢ Rainfall estimates can be computed for the entire planet by making use of both Polar Orbiting and than that associated with weighing gauges. Geostationary satellites. (National Weather Service, 2010) (National Weather Service, 2010)

23 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-C: Real Time Data Acquisition and Monitoring Network 24 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best 4.2 Flood-related weather observation in India

The real-time monitoring and statistical analysis of district wise daily rainfall is one of the important functions of the Hydro Meteorological Division of IMD at New Delhi. Based on the real time daily rainfall data, weekly district wise, sub-division wise and state wise rainfall distribution summaries are prepared regularly by the Rainfall Monitoring Unit. Maps showing weekly and cumulative rainfall figures in 36 meteorological subdivisions of the country are prepared. Hourly rainfall data in real time are available for numerous stations via the IMD website (http://www.imd.gov.in/section/hydro/QPF/WRF/).

Flood Met office, Hyderabad (Godavari) Inlt: 2012-11-25_00:00:00 Valld: 2012-11-26_03:00:00 IMD WRF Rainfall forecast (mm)

220 N 200

0 21 N 130

0 70 20 N

Indian Ocean 24-HOUR Rainfall Ending 12Z 40 Experimental NOAA/NESDIS Hydro- Estimator Product 190 N MM O 20 40 60 80 100 120 140 160 180 200 220 240 260 20

180 N 10 Figure 4.4: Example of an experimental product of 24 hours rainfall estimation from 1 170 N satellite data (http://www.star.nesdis.noaa.gov/)

160 N 740 E 760 E 780 E 800 E 820 E Satellite Precipitation Estimation Resources (MetEd, 2011):

Sub basin wise Average Rainfall Forecast ¢ Virtual Institute for Satellite Integration Training (VISIT) links and tutorials UPPER GODAVRI(G1) = 1mm MANJIRA (G4) = 1mm http://rammb.cira.colostate.edu/training/visit/links_and_tutorials/ LOWER GODAVRI(G10) = 2mm MIDDLE GODAVRI (G5) = 1mm INDRAVATHI (G 11) = 0mm MANERU (G6) = 1mm ¢ CMORPH: http://www.cpc.noaa.gov/products/janowiak/cmorph_description.html SABARI (G12) = 0mm PENGANGA (G7) = 2mm PRAVARA (G2) = 0mm WARDHA (G8)= 1mm ¢ HYDRO-ESTIMATOR: http://www.star.nesdis.noaa.gov/smcd/emb/ff/HydroEst.php PURNA (G3)= 0mm PRANAHITA/WAINGANGA (G9)= 0mm

OUTPUT FROM WRF 3. 1 MODEL ¢ EUMETSAT MPE: http://oiswww.eumetsat.org/IPPS/html/MSG/PRODUCTS/MPE/ WE=433; SN=433; Levels=38; Dis=9 km; Phys Opt=4; PBL Opt=2; Cu Opt=5

¢ PERSIANN: http://chrs.web.uci.edu/research/satellite_precipitation/activities00.html Figure 4.5: Quantitative precipitation forecasts, WRF ARW 9Km ´ 9Km (Source: India Meteorology Department) ¢ SCaMPR: http://www.star.nesdis.noaa.gov/smcd/emb/ff/SCaMPR.php

25 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-C: Real Time Data Acquisition and Monitoring Network 26 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best 4.2 Flood-related weather observation in India

The real-time monitoring and statistical analysis of district wise daily rainfall is one of the important functions of the Hydro Meteorological Division of IMD at New Delhi. Based on the real time daily rainfall data, weekly district wise, sub-division wise and state wise rainfall distribution summaries are prepared regularly by the Rainfall Monitoring Unit. Maps showing weekly and cumulative rainfall figures in 36 meteorological subdivisions of the country are prepared. Hourly rainfall data in real time are available for numerous stations via the IMD website (http://www.imd.gov.in/section/hydro/QPF/WRF/).

Flood Met office, Hyderabad (Godavari) Inlt: 2012-11-25_00:00:00 Valld: 2012-11-26_03:00:00 IMD WRF Rainfall forecast (mm)

220 N 200

0 21 N 130

0 70 20 N

Indian Ocean 24-HOUR Rainfall Ending 12Z 40 Experimental NOAA/NESDIS Hydro- Estimator Product 190 N MM O 20 40 60 80 100 120 140 160 180 200 220 240 260 20

180 N 10 Figure 4.4: Example of an experimental product of 24 hours rainfall estimation from 1 170 N satellite data (http://www.star.nesdis.noaa.gov/)

160 N 740 E 760 E 780 E 800 E 820 E Satellite Precipitation Estimation Resources (MetEd, 2011):

Sub basin wise Average Rainfall Forecast ¢ Virtual Institute for Satellite Integration Training (VISIT) links and tutorials UPPER GODAVRI(G1) = 1mm MANJIRA (G4) = 1mm http://rammb.cira.colostate.edu/training/visit/links_and_tutorials/ LOWER GODAVRI(G10) = 2mm MIDDLE GODAVRI (G5) = 1mm INDRAVATHI (G 11) = 0mm MANERU (G6) = 1mm ¢ CMORPH: http://www.cpc.noaa.gov/products/janowiak/cmorph_description.html SABARI (G12) = 0mm PENGANGA (G7) = 2mm PRAVARA (G2) = 0mm WARDHA (G8)= 1mm ¢ HYDRO-ESTIMATOR: http://www.star.nesdis.noaa.gov/smcd/emb/ff/HydroEst.php PURNA (G3)= 0mm PRANAHITA/WAINGANGA (G9)= 0mm

OUTPUT FROM WRF 3. 1 MODEL ¢ EUMETSAT MPE: http://oiswww.eumetsat.org/IPPS/html/MSG/PRODUCTS/MPE/ WE=433; SN=433; Levels=38; Dis=9 km; Phys Opt=4; PBL Opt=2; Cu Opt=5

¢ PERSIANN: http://chrs.web.uci.edu/research/satellite_precipitation/activities00.html Figure 4.5: Quantitative precipitation forecasts, WRF ARW 9Km ´ 9Km (Source: India Meteorology Department) ¢ SCaMPR: http://www.star.nesdis.noaa.gov/smcd/emb/ff/SCaMPR.php

25 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-C: Real Time Data Acquisition and Monitoring Network 26 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best Flood Meteorological Offices (FMOs) have been set up by IMD at ten locations viz., Agra, Ahmedabad, ¢ Jalpaiguri: Teesta Asansol, Bhubaneshwar, Guwahati, Hyderabad, Jalpaiguri, Lucknow, New Delhi and Patna. During the flood ¢ Lucknow: Ganga, Ramganga, Gomti, Sai, Rapti Ghagra and Samda season, FMOs provide valuable meteorological support to the Central Water Commission for issuing flood warnings in respect of the following rivers (from http://www.imd.gov.in/services/others/hydrometeorology- ¢ New Delhi: Upper Yamuna, Lower Yamuna, Sahibi flood.htm): ¢ Patna: Kosi, Mahananda, Baghmati, Kamla, Gandak, Buri Gandak, North Koel, Kanhar, Pun Pun and Upper Sone. ¢ Agra: Lower Yamuna and Betwa

¢ Ahmedabad: Narmada, Tapi, Mahi, Sabarmati, Banas and Deman Ganga For these river basins IMD also calculates quantitative precipitation forecasts based on the WRF model over 3 days (see chapter 5). ¢ Asansol: Ajay, Mayurakshi and Kangsabati

¢ Bhubaneshwar: Mahanadi, Brahmani, Baiterini, Bruhaba-lang, Subernarekha, Rushkulya and Vansdhara

¢ Guwahati: Brahmaputra and Barak

¢ Hyderabad: Godawari and Krishna KALPANA - 1 QPE(mm. 10 X10 ) 18AUG2012 DAILY

3DN DELHI Surface Rainfall R_150 KM 2DN Task : IMD–B PRF : 600/450 Melt : 5.5 km Max Range:150 Km 1DN 13:32:51Z 18 AUG 2012 UTC EQ

105> 113 98 105 90 98 10S 83 90 75 83 68 75 60 68 20S 53 60 45 53 38 45 30 38 30S

23 30 in mm/h Rate Rainfall 15 23 7.5 15 40S 0.0 7.5 30E 40E 50E 60E 70E 80E 90E 100E 110E 120E

1 1O 19 28 37 46 55 64 73 Figure 4.6: Surface rainfall round Delhi. The India Meteorological Department (IMD) issues on its website http://www.imd.gov.in/main_new.htm Doppler Radar Products of 14 locations Figure 4.7: Quantitative precipitation estimate from satellite data (IMD http://www.imd.gov.in) over India with surface rainfall intensity and precipitation accumulation over 24 hours.

27 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-C: Real Time Data Acquisition and Monitoring Network 28 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best Flood Meteorological Offices (FMOs) have been set up by IMD at ten locations viz., Agra, Ahmedabad, ¢ Jalpaiguri: Teesta Asansol, Bhubaneshwar, Guwahati, Hyderabad, Jalpaiguri, Lucknow, New Delhi and Patna. During the flood ¢ Lucknow: Ganga, Ramganga, Gomti, Sai, Rapti Ghagra and Samda season, FMOs provide valuable meteorological support to the Central Water Commission for issuing flood warnings in respect of the following rivers (from http://www.imd.gov.in/services/others/hydrometeorology- ¢ New Delhi: Upper Yamuna, Lower Yamuna, Sahibi flood.htm): ¢ Patna: Kosi, Mahananda, Baghmati, Kamla, Gandak, Buri Gandak, North Koel, Kanhar, Pun Pun and Upper Sone. ¢ Agra: Lower Yamuna and Betwa

¢ Ahmedabad: Narmada, Tapi, Mahi, Sabarmati, Banas and Deman Ganga For these river basins IMD also calculates quantitative precipitation forecasts based on the WRF model over 3 days (see chapter 5). ¢ Asansol: Ajay, Mayurakshi and Kangsabati

¢ Bhubaneshwar: Mahanadi, Brahmani, Baiterini, Bruhaba-lang, Subernarekha, Rushkulya and Vansdhara

¢ Guwahati: Brahmaputra and Barak

¢ Hyderabad: Godawari and Krishna KALPANA - 1 QPE(mm. 10 X10 ) 18AUG2012 DAILY

3DN DELHI Surface Rainfall R_150 KM 2DN Task : IMD–B PRF : 600/450 Melt : 5.5 km Max Range:150 Km 1DN 13:32:51Z 18 AUG 2012 UTC EQ

105> 113 98 105 90 98 10S 83 90 75 83 68 75 60 68 20S 53 60 45 53 38 45 30 38 30S

23 30 in mm/h Rate Rainfall 15 23 7.5 15 40S 0.0 7.5 30E 40E 50E 60E 70E 80E 90E 100E 110E 120E

1 1O 19 28 37 46 55 64 73 Figure 4.6: Surface rainfall round Delhi. The India Meteorological Department (IMD) issues on its website http://www.imd.gov.in/main_new.htm Doppler Radar Products of 14 locations Figure 4.7: Quantitative precipitation estimate from satellite data (IMD http://www.imd.gov.in) over India with surface rainfall intensity and precipitation accumulation over 24 hours.

27 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-C: Real Time Data Acquisition and Monitoring Network 28 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best Dependent from the lead time of the forecast there are: Examples of global models are the European Centre for Medium-Range Weather Forecasting (ECMWF) deterministic and ensemble prediction systems, the German Weather Service global model (GME) or the Global ¢ Now-casts for a few hours, Forecast System (GFS) run by NOAA. Many national meteorological agencies operate limited area models, for example, the ARPS University of Oklahoma, COSMO-DE (German Weather Service), HIRLAM (Danish ¢ Short-range forecasts between 4 and 48 hours Meteorological Institute) or the National Centre for Medium Range Weather Forecasting (NCMRWF) in India. ¢ Medium-range forecasts between 2 and 10 days and The numerical weather forecast products are the most important source for analysing the flood risk over the ¢ Long-range or seasonal forecasts for 10 to 365 days. next 2 to 5 days.

Now-casts are based on the extrapolation of observations in Obtaining a quantitative precipitation forecast (QPF) at appropriate spatial and temporal scales is not an easy time. Ground measurements of precipitation for instance can be 5 task as rainfall is one of the most difficult to forecast element of the hydrological cycle. This is especially true in forecasted by using time series analysis techniques, stochastic India with the indeterminacy of monsoon rains. LU-D: models or artificial neural networks. Animated pictures of observed radar rainfall often show a movement of rainfall fields Questions or frontal rains. This movement can be forecasted for the next What kinds of numerical weather forecast products exist? Meteorological few hours by keeping the velocity and the direction of the Who does it? propagation. The German weather service publishes a two hour From whom can I get it? now-cast based on this method. Forecasts What does it cost? For longer lead times, improvements of skill levels in numerical weather prediction models (NWP) made it more feasible to 5.1 German Weather Service (DWD) integrate weather prediction with flood forecasting systems. Numerical weather prediction models are available either as global models or as regional models (also known as limited-area With a grid length of 30 km on 60 vertical layers, the global model GME calculates at a total of 39 million grid models, or LAMS) for a particular section of the atmosphere. points the temporal evolution of weather parameters, such as air pressure, wind, water vapour, clouds and The latter are of a higher resolution and are nested in the global precipitation for up to seven days in advance. For Europe the regional model COSMO-EU (formerly known as models (Werner et al., 2006). LME), with a grid length of 7 km on 40 layers, uses a total of 17.5 million grid points to provide detailed weather forecasts for up to three days in advance. During computation the COSMO-EU receives the GME forecasts as lateral boundary values. The high resolution model COSMO-DE (formerly known as LMK), with a grid length of only 2.8 km and 50 layers at a total of 9.7 million grid points, has been delivering eighteen-hour forecasts for Germany 8 times per day, in particular for the warning of dangerous weather systems such as storms and thunderstorms. The lateral boundary values are derived from COSMO-EU forecasts.

The DWD provides the High Resolution Model (HRM) for use at other Meteorological Services, especially in developing and newly industrialised countries, such as Vietnam, the Philippines, or Kenya. GME forecast data are sent to these Meteorological Services twice a day via the Internet as lateral boundary values for calculations with the HRM (http://www.met.gov.om/hrm/index.php).

29 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-D: Meteorological Forecasts 30 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best Dependent from the lead time of the forecast there are: Examples of global models are the European Centre for Medium-Range Weather Forecasting (ECMWF) deterministic and ensemble prediction systems, the German Weather Service global model (GME) or the Global ¢ Now-casts for a few hours, Forecast System (GFS) run by NOAA. Many national meteorological agencies operate limited area models, for example, the ARPS University of Oklahoma, COSMO-DE (German Weather Service), HIRLAM (Danish ¢ Short-range forecasts between 4 and 48 hours Meteorological Institute) or the National Centre for Medium Range Weather Forecasting (NCMRWF) in India. ¢ Medium-range forecasts between 2 and 10 days and The numerical weather forecast products are the most important source for analysing the flood risk over the ¢ Long-range or seasonal forecasts for 10 to 365 days. next 2 to 5 days.

Now-casts are based on the extrapolation of observations in Obtaining a quantitative precipitation forecast (QPF) at appropriate spatial and temporal scales is not an easy time. Ground measurements of precipitation for instance can be 5 task as rainfall is one of the most difficult to forecast element of the hydrological cycle. This is especially true in forecasted by using time series analysis techniques, stochastic India with the indeterminacy of monsoon rains. LU-D: models or artificial neural networks. Animated pictures of observed radar rainfall often show a movement of rainfall fields Questions or frontal rains. This movement can be forecasted for the next What kinds of numerical weather forecast products exist? Meteorological few hours by keeping the velocity and the direction of the Who does it? propagation. The German weather service publishes a two hour From whom can I get it? now-cast based on this method. Forecasts What does it cost? For longer lead times, improvements of skill levels in numerical weather prediction models (NWP) made it more feasible to 5.1 German Weather Service (DWD) integrate weather prediction with flood forecasting systems. Numerical weather prediction models are available either as global models or as regional models (also known as limited-area With a grid length of 30 km on 60 vertical layers, the global model GME calculates at a total of 39 million grid models, or LAMS) for a particular section of the atmosphere. points the temporal evolution of weather parameters, such as air pressure, wind, water vapour, clouds and The latter are of a higher resolution and are nested in the global precipitation for up to seven days in advance. For Europe the regional model COSMO-EU (formerly known as models (Werner et al., 2006). LME), with a grid length of 7 km on 40 layers, uses a total of 17.5 million grid points to provide detailed weather forecasts for up to three days in advance. During computation the COSMO-EU receives the GME forecasts as lateral boundary values. The high resolution model COSMO-DE (formerly known as LMK), with a grid length of only 2.8 km and 50 layers at a total of 9.7 million grid points, has been delivering eighteen-hour forecasts for Germany 8 times per day, in particular for the warning of dangerous weather systems such as storms and thunderstorms. The lateral boundary values are derived from COSMO-EU forecasts.

The DWD provides the High Resolution Model (HRM) for use at other Meteorological Services, especially in developing and newly industrialised countries, such as Vietnam, the Philippines, or Kenya. GME forecast data are sent to these Meteorological Services twice a day via the Internet as lateral boundary values for calculations with the HRM (http://www.met.gov.om/hrm/index.php).

29 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-D: Meteorological Forecasts 30 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best 5.2 Numerical weather prediction in India GEFS : Rainfall (cm/day), Ens Mean 24–hr Forecast valid for OOZ18AUG2012 (IC = 00Z17AUG2012) 55N The India Meteorological Department publishes on its website (http://www.imd.gov.in) quantitative 50N precipitation forecasts based on the WRF model with a grid length of 9 km and a lead time of 3 days. In 45N addition there are basin rainfall estimate for different river basins in the area. 40N India Meteorological Department (IMD) has a mandate to issue short-range forecasts (validity of 2-3 days) and 35N long range seasonal forecast. National Centre for Medium Range Weather Forecasting (NCMRWF) has the 30N mandate for medium range (validity 4-10 days) forecasting. Many of the Numerical Models such as the 25N

Limited Area Model (LAM) and 5th Generation Meso-Scale Model (MM5) are being used for short range 20N

weather forecasts. IMD's long range forecast is based on statistical model that uses a number of global 15N

parameters called predictors which have high correlation with monsoon rainfall. At present, IMD uses 8 to 10 10N parameter models for long range forecast. In view of potential of numerical models, IMD has adopted an 5N experimental prediction system based on numerical models. For this purpose, IMD under a collaborative EQ research programme with Indian Institute of Science, Bangalore has adopted a numerical model developed at 5S the Experimental Climate Prediction Centre (ECPC), Scripps Institute of Oceanography, USA. 10S

15S The model resolution and the accuracy have limitations vis-à-vis the more advanced countries because of the 20E 30E 40E 50E 60E 70E 80E 90E 100E 110E 120E 130E 140E inadequate infrastructure such as observational network, computing capacity and human resources. Upgrade 0.1 1 2 4 8 16 32 64 of observational network, enhancement of computing facility for running high resolution numerical weather prediction models and improvement of communication network are required to meet the present requirements LEVELS 24 Hrs 48 Hrs 72 Hrs 96 Hrs 120 Hrs 144 Hrs 168 Hrs 192 Hrs 216 Hrs 240 Hrs and to achieve global standards in weather forecasting (http://www.dst.gov.in/admin_finance/sq288.htm, RAIN Ch-01 Ch-02 Ch-03 Ch-04 Ch-05 Ch-06 Ch-07 Ch-08 Ch-09 Ch-10 2006). Figure 5.1: Example of 24-hr precipitation forecast (http://www.ncmrwf.gov.in/) In India the National Centre for Medium Range Weather Forecasting (NCMRWF, http://www.ncmrwf.gov.in/) develops advanced numerical weather prediction systems, with increased reliability and accuracy over India. The National Centre for Medium Range Weather Forecasting (NCMRWF) makes available its various operational forecast products from its Global Forecast System at various resolutions on its ftp site with free access. These data will be retained in the ftp site only for a limited period. It is expected that the user community will acknowledge the data source used for the research in its publications and intimate us about the usage details which will motivate us to continue with this service and possibly inform the potential users regarding any changes in the data supply in advance.

31 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-D: Meteorological Forecasts 32 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best 5.2 Numerical weather prediction in India GEFS : Rainfall (cm/day), Ens Mean 24–hr Forecast valid for OOZ18AUG2012 (IC = 00Z17AUG2012) 55N The India Meteorological Department publishes on its website (http://www.imd.gov.in) quantitative 50N precipitation forecasts based on the WRF model with a grid length of 9 km and a lead time of 3 days. In 45N addition there are basin rainfall estimate for different river basins in the area. 40N India Meteorological Department (IMD) has a mandate to issue short-range forecasts (validity of 2-3 days) and 35N long range seasonal forecast. National Centre for Medium Range Weather Forecasting (NCMRWF) has the 30N mandate for medium range (validity 4-10 days) forecasting. Many of the Numerical Models such as the 25N

Limited Area Model (LAM) and 5th Generation Meso-Scale Model (MM5) are being used for short range 20N weather forecasts. IMD's long range forecast is based on statistical model that uses a number of global 15N parameters called predictors which have high correlation with monsoon rainfall. At present, IMD uses 8 to 10 10N parameter models for long range forecast. In view of potential of numerical models, IMD has adopted an 5N experimental prediction system based on numerical models. For this purpose, IMD under a collaborative EQ research programme with Indian Institute of Science, Bangalore has adopted a numerical model developed at 5S the Experimental Climate Prediction Centre (ECPC), Scripps Institute of Oceanography, USA. 10S

15S The model resolution and the accuracy have limitations vis-à-vis the more advanced countries because of the 20E 30E 40E 50E 60E 70E 80E 90E 100E 110E 120E 130E 140E inadequate infrastructure such as observational network, computing capacity and human resources. Upgrade 0.1 1 2 4 8 16 32 64 of observational network, enhancement of computing facility for running high resolution numerical weather prediction models and improvement of communication network are required to meet the present requirements LEVELS 24 Hrs 48 Hrs 72 Hrs 96 Hrs 120 Hrs 144 Hrs 168 Hrs 192 Hrs 216 Hrs 240 Hrs and to achieve global standards in weather forecasting (http://www.dst.gov.in/admin_finance/sq288.htm, RAIN Ch-01 Ch-02 Ch-03 Ch-04 Ch-05 Ch-06 Ch-07 Ch-08 Ch-09 Ch-10 2006). Figure 5.1: Example of 24-hr precipitation forecast (http://www.ncmrwf.gov.in/) In India the National Centre for Medium Range Weather Forecasting (NCMRWF, http://www.ncmrwf.gov.in/) develops advanced numerical weather prediction systems, with increased reliability and accuracy over India. The National Centre for Medium Range Weather Forecasting (NCMRWF) makes available its various operational forecast products from its Global Forecast System at various resolutions on its ftp site with free access. These data will be retained in the ftp site only for a limited period. It is expected that the user community will acknowledge the data source used for the research in its publications and intimate us about the usage details which will motivate us to continue with this service and possibly inform the potential users regarding any changes in the data supply in advance.

31 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-D: Meteorological Forecasts 32 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best The basic elements of the flood forecast system are: Dependent from the lead time of the forecast we distinguish:

¢ ¢ real time data acquisition, Now-casts for a few hours,

¢ Short-range forecasts between 4 and 48 hours, ¢ hydrologic and hydraulic models, ¢ Medium-range forecasts between 2 and 10 days, and ¢ update of the forecast and data assimilation. ¢ Long-range or seasonal forecasts for 10 to 365 days. Real time data acquisition delivers the necessary input data to the hydrologic and hydraulic models like precipitation data or upstream The origin of these types is the weather forecast and it works for the flood forecast too. Long-range forecasts 6 discharge data. Hydrologic and hydraulic models simulate hydrologic are of little practical use in flood warning. Medium-range flood forecasts are based on medium-range weather flood processes and calculate forecasts. A wide range of hydrologic forecasts with the same degree of reliability. There may be travel times of the flood wave of several days in LU-E: and hydraulic modelling approaches may be applied where models large rivers. In this case even medium range flood forecasts can be based on measured upstream gauging take information on the current and past states of the system, and stations with a high degree of reliability. forecasts are made for the desired lead time as a function of Flood boundary inputs on the system. Where necessary, forecasts of With forecast models we get measures like water level and discharge usually at a gauging station. Mostly, the meteorological conditions are required to allow issue of warnings at expected flood peak is of highest interest. But there are other quantities, like Forecast sufficient lead time. Generally, these forecasts are sourced from meteorological services, and must be integrated within the flood ¢ full hydrograph (i.e. to estimate the expected inflow volume to a reservoir), forecasting system. Through the process of data assimilation and ¢ System updating, simulated data are combined with real-time data to provide probability of overtopping certain critical flood stages (i.e. to manage flood protection response), a more accurate forecast and to equal the last observation point to ¢ duration of overtopping a critical flood stage, and the first forecast point. The use of an updating or data assimilation technique is an important aspect of applying models in real time ¢ forecast of the expected flooded area. (Werner et al., 2006).

Accuracy, reliability and timeliness are basic requirements in 6.1 Hydrologic response time and lead time of the forecast operation of the forecast system. Real time operational forecast is very much a time critical exercise (Szöllösi-Nagy, A.,2009). The difference between the forecast of a meteorological event like precipitation and the hydrological event like Simplicity of the methods used to process the input data and to a flood peak is the dependence of the forecast-lead time from the catchment size. calculate the forecasts is a key requirement. Robustness of the system components i.e. the possibility to handle with missing input Preliminary considerations about the flood drainage behaviour of the catchment at the forecast point are data in real time is another main pillar. essential to decide which type of forecast models and input data are necessary to get a reliable forecast. In particular, observations of the travel time of the flood peak, the time to peak and the time of runoff concentration in the catchment provide information about reachable forecast lead times specially without extension by less accurate precipitation forecasts.

33 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-E: Flood Forecast System 34 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best The basic elements of the flood forecast system are: Dependent from the lead time of the forecast we distinguish:

¢ ¢ real time data acquisition, Now-casts for a few hours,

¢ Short-range forecasts between 4 and 48 hours, ¢ hydrologic and hydraulic models, ¢ Medium-range forecasts between 2 and 10 days, and ¢ update of the forecast and data assimilation. ¢ Long-range or seasonal forecasts for 10 to 365 days. Real time data acquisition delivers the necessary input data to the hydrologic and hydraulic models like precipitation data or upstream The origin of these types is the weather forecast and it works for the flood forecast too. Long-range forecasts 6 discharge data. Hydrologic and hydraulic models simulate hydrologic are of little practical use in flood warning. Medium-range flood forecasts are based on medium-range weather flood processes and calculate forecasts. A wide range of hydrologic forecasts with the same degree of reliability. There may be travel times of the flood wave of several days in LU-E: and hydraulic modelling approaches may be applied where models large rivers. In this case even medium range flood forecasts can be based on measured upstream gauging take information on the current and past states of the system, and stations with a high degree of reliability. forecasts are made for the desired lead time as a function of Flood boundary inputs on the system. Where necessary, forecasts of With forecast models we get measures like water level and discharge usually at a gauging station. Mostly, the meteorological conditions are required to allow issue of warnings at expected flood peak is of highest interest. But there are other quantities, like Forecast sufficient lead time. Generally, these forecasts are sourced from meteorological services, and must be integrated within the flood ¢ full hydrograph (i.e. to estimate the expected inflow volume to a reservoir), forecasting system. Through the process of data assimilation and ¢ System updating, simulated data are combined with real-time data to provide probability of overtopping certain critical flood stages (i.e. to manage flood protection response), a more accurate forecast and to equal the last observation point to ¢ duration of overtopping a critical flood stage, and the first forecast point. The use of an updating or data assimilation technique is an important aspect of applying models in real time ¢ forecast of the expected flooded area. (Werner et al., 2006).

Accuracy, reliability and timeliness are basic requirements in 6.1 Hydrologic response time and lead time of the forecast operation of the forecast system. Real time operational forecast is very much a time critical exercise (Szöllösi-Nagy, A.,2009). The difference between the forecast of a meteorological event like precipitation and the hydrological event like Simplicity of the methods used to process the input data and to a flood peak is the dependence of the forecast-lead time from the catchment size. calculate the forecasts is a key requirement. Robustness of the system components i.e. the possibility to handle with missing input Preliminary considerations about the flood drainage behaviour of the catchment at the forecast point are data in real time is another main pillar. essential to decide which type of forecast models and input data are necessary to get a reliable forecast. In particular, observations of the travel time of the flood peak, the time to peak and the time of runoff concentration in the catchment provide information about reachable forecast lead times specially without extension by less accurate precipitation forecasts.

33 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-E: Flood Forecast System 34 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best There are three different types (simplified approach from Werner et al. 2006): One example should clarify this issue (for type 1):

¢ Type 1: The time the water needs to flow through the river channel is greater than the time water needs to Consider the Nile River coming from the highlands of Ethiopia: In the time of heavy rains there, large amounts flow as runoff over land. The flood routing process dominates form and volume of the flood wave. Forecasts of water start to flow, but on the way through Egypt with a desert-like condition, no additional water reaches can be based on upstream gauging measurements. the Nile and forecast lead times depends only from the travel time of the peak along the river. Here the challenge is modelling the loss of water by irrigation. ¢ Type 2: The time the water flows over land to the river is greater than the time the water flows through the river channel, concentration of runoff dominates form and volume of the flood hydrograph. The possible lead time is connected to the time to peak or the time between the rainfall event and the flood peak. Questions Forecasts are based on precipitation and/or snow melt measurements. What are the three main components of a flood forecast system?

¢ Type 3: Desired lead time is greater than the time the water needs to flow over the land and through the What is the hydrological response time and how it is related to the lead time of the forecast? river channel. It means that the response time of the system to the rain or to the snow melt is shorter than What are three basic requirements to a forecast system? the desired lead time of the forecast and we need a forecast of rain or snow melt to get a sufficient lead What are the four main lead time horizons? time. This type is found in small catchments and flash flood situations.

Forecast point Catchment boundary Main river

Figure 6.1: Scheme of catchment with forecast points I – VII and sub basins A - F (Werner et al., 2006)

Exercise: Assign type 1, 2 or 3 to the forecast point I, II, III, IV, V, VI and VII in Figure 6.1

35 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-E: Flood Forecast System 36 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best There are three different types (simplified approach from Werner et al. 2006): One example should clarify this issue (for type 1):

¢ Type 1: The time the water needs to flow through the river channel is greater than the time water needs to Consider the Nile River coming from the highlands of Ethiopia: In the time of heavy rains there, large amounts flow as runoff over land. The flood routing process dominates form and volume of the flood wave. Forecasts of water start to flow, but on the way through Egypt with a desert-like condition, no additional water reaches can be based on upstream gauging measurements. the Nile and forecast lead times depends only from the travel time of the peak along the river. Here the challenge is modelling the loss of water by irrigation. ¢ Type 2: The time the water flows over land to the river is greater than the time the water flows through the river channel, concentration of runoff dominates form and volume of the flood hydrograph. The possible lead time is connected to the time to peak or the time between the rainfall event and the flood peak. Questions Forecasts are based on precipitation and/or snow melt measurements. What are the three main components of a flood forecast system?

¢ Type 3: Desired lead time is greater than the time the water needs to flow over the land and through the What is the hydrological response time and how it is related to the lead time of the forecast? river channel. It means that the response time of the system to the rain or to the snow melt is shorter than What are three basic requirements to a forecast system? the desired lead time of the forecast and we need a forecast of rain or snow melt to get a sufficient lead What are the four main lead time horizons? time. This type is found in small catchments and flash flood situations.

Forecast point Catchment boundary Main river Tributary

Figure 6.1: Scheme of catchment with forecast points I – VII and sub basins A - F (Werner et al., 2006)

Exercise: Assign type 1, 2 or 3 to the forecast point I, II, III, IV, V, VI and VII in Figure 6.1

35 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-E: Flood Forecast System 36 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best The building of models in makes use of methods of applied General Components of a Complex Hydrologic Model System system analysis. System analysis deals with the analysis of complex System input systems. The objective of this analysis is the construction of a model System inflow Storage as a simplified and abstract image of the real system. Movement of surface or subsurface water System outflow Precipitation System outflows System outflow The models in hydrology are based on the physical law of mass Rain Sleet Hail Snow Basin channel Evaporation outflow conservation (continuity). The sum of changes in input and output of System outflow the system has to be the sum of changes in the state of the system at Snow accumulation Evaporation Interception and melt System a certain time step. Inputs to a system are transferred to outputs and outflow Diversions may be calculated by a mathematical transfer function below. System Depression Water on Surface Channel outflow 7 storage land surface runoff flow Evaporation

LU-F: Infiltration Channel seepage Reservoir storage

System Root zone Subsurface Water allocation System Hydrologic outflow System outflow storage stormflow Input p(t) transfer function: Output q(t) U {p(t)} Channel Evapotranspiration Percolation System and outflow

Imported Groundwater Groundwater Groundwater Groundwater Hydraulic water inflow storage flow outflow basin Input – Output – System with q(t) = U {p(t)}, where U is the System outflow system- or transfer function Models System inflow System inflow The transfer function describes the transformation of the input (i.e. Figure 7.1: Complex hydrologic model system (MetEd, 2010) precipitation) to the output (i.e. runoff). A simple case of a transfer function is a linear regression function as shown in Figure 7.2. 7.1 Black box model The typical system in hydrologic forecasts may be a catchment with input of precipitation and output of runoff or a channel reach with In a black box model input and output values are determined only by measuring data and expressed in inflow and outflow. mathematical algorithms. The transfer function is determined without any knowledge of the internal working of the system only by its input and output measurements. Different models of this type can be distinguished by The system can be composed of different components like the the type of the transfer function. These can be correlation functions (Figure 7.2), filtering techniques known complex hydrologic model system in Figure 7.1. from signal processing and time series analysis or simple conceptual models like the unit hydrograph.

There are different models described briefly in Table 7.1. Methods of data mining, artificial neural networks or genetic programming are further methods to formulate the transfer function without an explicit physical description of the system.

37 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-F: Hydrologic and Hydraulic Models 38 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best The building of models in hydrology makes use of methods of applied General Components of a Complex Hydrologic Model System system analysis. System analysis deals with the analysis of complex System input systems. The objective of this analysis is the construction of a model System inflow Storage as a simplified and abstract image of the real system. Movement of surface or subsurface water System outflow Precipitation System outflows System outflow The models in hydrology are based on the physical law of mass Rain Sleet Hail Snow Basin channel Evaporation outflow conservation (continuity). The sum of changes in input and output of System outflow the system has to be the sum of changes in the state of the system at Snow accumulation Evaporation Interception and melt System a certain time step. Inputs to a system are transferred to outputs and outflow Diversions may be calculated by a mathematical transfer function below. System Depression Water on Surface Channel outflow 7 storage land surface runoff flow Evaporation

LU-F: Infiltration Channel seepage Reservoir storage

System Root zone Subsurface Water allocation System Hydrologic outflow System outflow storage stormflow Input p(t) transfer function: Output q(t) U {p(t)} Channel Evapotranspiration Percolation System and baseflow outflow

Imported Groundwater Groundwater Groundwater Groundwater Hydraulic water inflow storage flow outflow basin Input – Output – System with q(t) = U {p(t)}, where U is the System outflow system- or transfer function Models System inflow System inflow The transfer function describes the transformation of the input (i.e. Figure 7.1: Complex hydrologic model system (MetEd, 2010) precipitation) to the output (i.e. runoff). A simple case of a transfer function is a linear regression function as shown in Figure 7.2. 7.1 Black box model The typical system in hydrologic forecasts may be a catchment with input of precipitation and output of runoff or a channel reach with In a black box model input and output values are determined only by measuring data and expressed in inflow and outflow. mathematical algorithms. The transfer function is determined without any knowledge of the internal working of the system only by its input and output measurements. Different models of this type can be distinguished by The system can be composed of different components like the the type of the transfer function. These can be correlation functions (Figure 7.2), filtering techniques known complex hydrologic model system in Figure 7.1. from signal processing and time series analysis or simple conceptual models like the unit hydrograph.

There are different models described briefly in Table 7.1. Methods of data mining, artificial neural networks or genetic programming are further methods to formulate the transfer function without an explicit physical description of the system.

37 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-F: Hydrologic and Hydraulic Models 38 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best The main attraction of black box model is that the model structure is identified using available data only. This 7.2 Distributed models leads to efficiently parameterized model structures. These are also relatively simple to implement and computationally very efficient. Computer technology made it possible to t apply more physically based and spatially However the lead time is limited to a time-lag between input event and output event. By applying the transfer I distributed models in flood forecast. In most function to the flood routing process the lead time corresponds to the travel time of the flood wave. Applied to forecasting centres in Europe this type of Q the process of runoff concentration it is in the order of the response time of the catchment to precipitation. models are used.

Because there is no physical description of the system a change in the physical properties leads to invalid In distributed models the catchment is t results and the historic database can no longer be used to identify the transfer function. Q divided in smaller subunits. The subdivision runoff hydrograph may be done by a regular grid of cells or by

Multiple linear regression HQ- peaks of gauge Wiblingen in relation to HQ-peaks of gauge Kempten 3-days-sum small sub catchments using geographical of precipitation station Illertissen based on 25 events from 1980 to 1993 information system. The size of the grids or sub basins varies with the size of the spatial Q Wibl. - 0,965* Q Kempt. + 0,922* P Illert. + 38,377 Standard error: 31,8 m3 /s multiple Corr. Cofficient.: 0,870 scale of the processes, the scale of available 1050 geographic maps and the computational 1000 200 160 possibilities. Most common is a grid size of 950 120 900 80 1x1km. For very large catchments with areas 850 40 of more than 50000 km² often a grid size of 0 Figure 7.3: Scheme of the unit-hydrograph concept 800 4x4 or 5x5 km are used.

3 (Mauser, 2012), the hydrograph response to a unit 750 rainfall (unit hydrograph) is super positioned by 700 convolution according to the measured rainfall 650 intensities (I) to get the runoff hydrograph (Q). 600

HQ Wiblingen [mHQ Wiblingen /s] 550 500 peak travel time Kempten - Wiblingen: 450 10 +/ -3 Stunden 400 350

300 3-day precip. sum Illertissen [mm] 260 280 300 320 340 360 380 400 420 440 460 480 500 520 540 560 580 600 620 640 660 680 700 720 740 760 780 800 HQ Kempten [m3 /s]

Figure 7.2: Example of a multiple-linear correlation model relating peak discharge to an upstream gauge site and to the precipitation sum representing additional inflow between the gauge sites (own design).

Figure 7.4: Division by sub-basins

39 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-F: Hydrologic and Hydraulic Models 40 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best The main attraction of black box model is that the model structure is identified using available data only. This 7.2 Distributed models leads to efficiently parameterized model structures. These are also relatively simple to implement and computationally very efficient. Computer technology made it possible to t apply more physically based and spatially However the lead time is limited to a time-lag between input event and output event. By applying the transfer I distributed models in flood forecast. In most function to the flood routing process the lead time corresponds to the travel time of the flood wave. Applied to forecasting centres in Europe this type of Q the process of runoff concentration it is in the order of the response time of the catchment to precipitation. models are used.

Because there is no physical description of the system a change in the physical properties leads to invalid In distributed models the catchment is t results and the historic database can no longer be used to identify the transfer function. Q divided in smaller subunits. The subdivision runoff hydrograph may be done by a regular grid of cells or by

Multiple linear regression HQ- peaks of gauge Wiblingen in relation to HQ-peaks of gauge Kempten 3-days-sum small sub catchments using geographical of precipitation station Illertissen based on 25 events from 1980 to 1993 information system. The size of the grids or sub basins varies with the size of the spatial Q Wibl. - 0,965* Q Kempt. + 0,922* P Illert. + 38,377 Standard error: 31,8 m3 /s multiple Corr. Cofficient.: 0,870 scale of the processes, the scale of available 1050 geographic maps and the computational 1000 200 160 possibilities. Most common is a grid size of 950 120 900 80 1x1km. For very large catchments with areas 850 40 of more than 50000 km² often a grid size of 0 Figure 7.3: Scheme of the unit-hydrograph concept 800 4x4 or 5x5 km are used.

3 (Mauser, 2012), the hydrograph response to a unit 750 rainfall (unit hydrograph) is super positioned by 700 convolution according to the measured rainfall 650 intensities (I) to get the runoff hydrograph (Q). 600

HQ Wiblingen [mHQ Wiblingen /s] 550 500 peak travel time Kempten - Wiblingen: 450 10 +/ -3 Stunden 400 350

300 3-day precip. sum Illertissen [mm] 260 280 300 320 340 360 380 400 420 440 460 480 500 520 540 560 580 600 620 640 660 680 700 720 740 760 780 800 HQ Kempten [m3 /s]

Figure 7.2: Example of a multiple-linear correlation model relating peak discharge to an upstream gauge site and to the precipitation sum representing additional inflow between the gauge sites (own design).

Figure 7.4: Division by sub-basins

39 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-F: Hydrologic and Hydraulic Models 40 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Table 7.1: Model types and characteristics Best

Examples Hydrologic Geographical (see box 1 Model Type Concepts and Input Data Advantages Disadvantages Data below for processes details)

Block Model Rainfall – Precipitation No Simple to Limited lead Regression, Runoff in or upstream implement time, valid filter flood flow data and to only in (Wiener, situation or calculate observed Kalman), flood routing range used for ARIMA, calibration ARMAX

Distributed Direct runoff, Precipitation, Geographical More Estimation of FGMOD event based rainfall loss (Air data of sub physically losses to get NAM model (snow melt) temperature basins or based and the volume of and wind grids and spatially direct runoff, speed for channels distributed less physically snow melt) based

Distributed Interception, Precipitation, Geographical More Big number of HBV water Infiltration, potential data of sub physically input HEC-HMS Figure 7.5: Division by regular grids balance Percolation, evaporation, basins or based and variables, high TOPKAPI model Runoff, wind, air grids, spatially skill for LARSIM evapotranspir pressure, channels, distributed, calibration SHE For each cell or sub basin the vertical fluxes and storage of water (artificial and manmade ponds and tanks, ation, vapour soil, land better NWSRFS lakes, storage in snow compaction, snow melt, interception, depression storage, infiltration, percolation, soil components pressure use, digital estimation of LISFLOOD storage, groundwater recharge, evaporation and transpiration) are calculated depending on the land use and of surface, elevation saturation soil and model state in the properties of soil. Movement of surface flow and subsurface flow are calculated either by simplified concepts of groundwater basin translation and linear storage, the concept of parallel linear cascades or by the more physically based kinematic wave method for surface flow and the Richards equation for soil water movement. Hydraulic 1- Water Water level Length, slope Physically HEC-RAS D Model movement and/or and cross based, short MIKE 11 and water discharge section of computing SOBEK Regarding the vertical water fluxes there are two types of models. First, the event based models considers only depth in the channels and times the most dominant processes during flood (rainfall and summarized water losses due to interception and floodplain infiltration) in a very simple concept. Water losses by evapo-transpiration are neglected. At the beginning of the Hydraulic 2- Water Water level Digital Physically Long SOBEK event a factor can be applied to the rate of direct runoff volume depending on the water saturation state of the D Model movement and/or elevation based, computing MIKE 21 catchment. and water discharge model of the complex flow times prevent depth in the floodplain in situation, application in Second the water balance model accounts for all processes of the water cycle over land including evapo- floodplain, high modelling flood forecast backwater resolution flood extent transpiration and soil moisture accounting (Figure 7.1). effects

41 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-F: Hydrologic and Hydraulic Models 42 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Table 7.1: Model types and characteristics Best

Examples Hydrologic Geographical (see box 1 Model Type Concepts and Input Data Advantages Disadvantages Data below for processes details)

Block Model Rainfall – Precipitation No Simple to Limited lead Regression, Runoff in or upstream implement time, valid filter flood flow data and to only in (Wiener, situation or calculate observed Kalman), flood routing range used for ARIMA, calibration ARMAX

Distributed Direct runoff, Precipitation, Geographical More Estimation of FGMOD event based rainfall loss (Air data of sub physically losses to get NAM model (snow melt) temperature basins or based and the volume of and wind grids and spatially direct runoff, speed for channels distributed less physically snow melt) based

Distributed Interception, Precipitation, Geographical More Big number of HBV water Infiltration, potential data of sub physically input HEC-HMS Figure 7.5: Division by regular grids balance Percolation, evaporation, basins or based and variables, high TOPKAPI model Runoff, wind, air grids, spatially skill for LARSIM evapotranspir pressure, channels, distributed, calibration SHE For each cell or sub basin the vertical fluxes and storage of water (artificial and manmade ponds and tanks, ation, vapour soil, land better NWSRFS lakes, storage in snow compaction, snow melt, interception, depression storage, infiltration, percolation, soil components pressure use, digital estimation of LISFLOOD storage, groundwater recharge, evaporation and transpiration) are calculated depending on the land use and of surface, elevation saturation soil and model state in the properties of soil. Movement of surface flow and subsurface flow are calculated either by simplified concepts of groundwater basin translation and linear storage, the concept of parallel linear cascades or by the more physically based kinematic wave method for surface flow and the Richards equation for soil water movement. Hydraulic 1- Water Water level Length, slope Physically HEC-RAS D Model movement and/or and cross based, short MIKE 11 and water discharge section of computing SOBEK Regarding the vertical water fluxes there are two types of models. First, the event based models considers only depth in the channels and times the most dominant processes during flood (rainfall and summarized water losses due to interception and floodplain floodplain infiltration) in a very simple concept. Water losses by evapo-transpiration are neglected. At the beginning of the Hydraulic 2- Water Water level Digital Physically Long SOBEK event a factor can be applied to the rate of direct runoff volume depending on the water saturation state of the D Model movement and/or elevation based, computing MIKE 21 catchment. and water discharge model of the complex flow times prevent depth in the floodplain in situation, application in Second the water balance model accounts for all processes of the water cycle over land including evapo- floodplain, high modelling flood forecast backwater resolution flood extent transpiration and soil moisture accounting (Figure 7.1). effects

41 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-F: Hydrologic and Hydraulic Models 42 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best Box 1 7.3 The LARSIM model as an example for a distributed conceptual FGMOD is part of the LARSIM model, where it is possible to switch between event-based FGMOD mode model and the water balance mode of LARSIM, used in forecast centres of Germany, non-commercial product.

NAM by Danish Hydraulic Institute, the rainfall- comes with the hydrodynamic Model MIKE Vertical water storage / transport per lad use in the sub basin 11, commercial product. INTERCEPTION and SNOW STORAGE HBV-96 by Swedish Meteorological and Hydrological Institute used in many forecast centres in Sweden, EVAPOTRANSPIRATION Netherlands, Germany, Austria and also in India and in other countries, non-commercial product. http://www.smhi.se/sgn0106/if/hydrologi/hbv.htm. SOIL STORAGE HEC-HMS by Hydrologic Engineering Centre of the U.S. Army Corps of Engineers, the Hydrologic Modelling System used mainly in the U.S., non-commercial product. % of water Lateral Vertical saturated areas drainage perkolation TOPKAPI by the University of Bologna, Italy, used in the forecast centre for the Po River in Italy, non- commercial product.

LARSIM (Large Area Simulation Model) based on FGMOD by the University of Hannover (Germany), used in forecast centres in Germany, France and Austria, non-commercial product. CATCHMENT STORAGE (lateral water storage / transport in the sub basin) SHE (Système Hydrologique Européen) was originally developed by the Danish Hydraulic Institute (DHI), SOGREAH (France) and the Institute of Hydrology (UK), non-commercial product but as MIKE-SHE in Surface runoff Interflow Ground water connection with MIKE 11, commercial product by DHI storage storage storage

US National Weather Service US National Weather Service River Forecast System NWSRFS River Forecast System by the U.S. National Weather Service uses different model components, i.e. SAC-SMA (Sacramento Soil Moisture Accounting Model) used in forecast centres of the National Weather Service, non-commercial product. RIVERS AND LAKES per river section LISFLOOD by Joint Research Centre of the European Commission, used by EFAS (European Flood Alert System), non-commercial product, http://floods.jrc.ec.europa.eu/lisflood-model. Flood routing Branching out, Lack retention, HEC-RAS by Hydrologic Engineering Centre of the U.S. Army Corps of Engineers, 1D-hydrodynamic in the river flowing into reservoir control modelling of open channel flow, used in many countries, non-commercial product.

MIKE 11 and MIKE 21 by Danish Hydraulic Institute, commercial product. MIKE 11 offers a unique possibility to simulate catchment runoff in many ways, from simple empirical rainfall-runoff methods to complex fully distributed process-based oriented modelling with MIKE SHE and is used extensively in India by NIH, CWC etc. particularly under Hydrology Project –II by Ministry of Water Resources and also by several academic and research institutions. Runoff SOBEK by Delft Hydraulics and Deltares, used in forecast centres of Germany, Netherlands and others, commercial product. Figure 7.6: Scheme of the LARSIM model (Ludwig/Bremicker, 2007)

43 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-F: Hydrologic and Hydraulic Models 44 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best Box 1 7.3 The LARSIM model as an example for a distributed conceptual FGMOD is part of the LARSIM model, where it is possible to switch between event-based FGMOD mode model and the water balance mode of LARSIM, used in forecast centres of Germany, non-commercial product.

NAM by Danish Hydraulic Institute, the rainfall-runoff model comes with the hydrodynamic Model MIKE Vertical water storage / transport per lad use in the sub basin 11, commercial product. INTERCEPTION and SNOW STORAGE HBV-96 by Swedish Meteorological and Hydrological Institute used in many forecast centres in Sweden, EVAPOTRANSPIRATION Netherlands, Germany, Austria and also in India and in other countries, non-commercial product. http://www.smhi.se/sgn0106/if/hydrologi/hbv.htm. SOIL STORAGE HEC-HMS by Hydrologic Engineering Centre of the U.S. Army Corps of Engineers, the Hydrologic Modelling System used mainly in the U.S., non-commercial product. % of water Lateral Vertical saturated areas drainage perkolation TOPKAPI by the University of Bologna, Italy, used in the forecast centre for the Po River in Italy, non- commercial product.

LARSIM (Large Area Simulation Model) based on FGMOD by the University of Hannover (Germany), used in forecast centres in Germany, France and Austria, non-commercial product. CATCHMENT STORAGE (lateral water storage / transport in the sub basin) SHE (Système Hydrologique Européen) was originally developed by the Danish Hydraulic Institute (DHI), SOGREAH (France) and the Institute of Hydrology (UK), non-commercial product but as MIKE-SHE in Surface runoff Interflow Ground water connection with MIKE 11, commercial product by DHI storage storage storage

US National Weather Service US National Weather Service River Forecast System NWSRFS River Forecast System by the U.S. National Weather Service uses different model components, i.e. SAC-SMA (Sacramento Soil Moisture Accounting Model) used in forecast centres of the National Weather Service, non-commercial product. RIVERS AND LAKES per river section LISFLOOD by Joint Research Centre of the European Commission, used by EFAS (European Flood Alert System), non-commercial product, http://floods.jrc.ec.europa.eu/lisflood-model. Flood routing Branching out, Lack retention, HEC-RAS by Hydrologic Engineering Centre of the U.S. Army Corps of Engineers, 1D-hydrodynamic in the river flowing into reservoir control modelling of open channel flow, used in many countries, non-commercial product.

MIKE 11 and MIKE 21 by Danish Hydraulic Institute, commercial product. MIKE 11 offers a unique possibility to simulate catchment runoff in many ways, from simple empirical rainfall-runoff methods to complex fully distributed process-based oriented modelling with MIKE SHE and is used extensively in India by NIH, CWC etc. particularly under Hydrology Project –II by Ministry of Water Resources and also by several academic and research institutions. Runoff SOBEK by Delft Hydraulics and Deltares, used in forecast centres of Germany, Netherlands and others, commercial product. Figure 7.6: Scheme of the LARSIM model (Ludwig/Bremicker, 2007)

43 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-F: Hydrologic and Hydraulic Models 44 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best LARSIM includes mathematical computation procedures to continuously describe and quantify the spatial and Surface or subsurface water flowing off of a grid cell may either flow into a channel or, if no channels exist, temporal distribution of the water balance's essential land phase components like precipitation, evapo- into the neighbouring down-slope grid cell. Therefore the grid cells have to be hydrologically connected in the transpiration, infiltration, water storage and runoff. The water balance model LARSIM (Large Area Runoff model. In other words, water flowing from one grid cell flows to the downs-lope grid cells and so on until the Simulation Model) is a so called conceptual model, i.e. the complex processes in the natural system are outlet of the watershed is reached (Figure 8.1). reproduced by simplified model concepts. Flood routing (chapter 2.5 “Channel flow and flood routing”, Figure 2.6) in the channel can be calculated by Because rain is the input into the model it shouldn’t matter if the rain is caused by rain types of temperate the Williams method of flood routing or by a simplified translation and linear storage concept, where the climate or the monsoon type. Therefore the LARSIM model can be used in India (as well as the other parameters are dependent from the roughness of the channel bed. Alternatively the inflows to the channels can hydrologic models mentioned in this chapter). be stored and flood routing can be calculated by separate hydrodynamic models using the results of LARSIM as lateral inflow to the channel. Among others the following hydrological processes are considered in LARSIM: interception, evapo- transpiration, snow accumulation, snow compaction and snow melt, soil water storage as well as storage and lateral transport in streams and lakes (Figure 7.6).

The geographical information base for the LARSIM model includes a digital terrain model, the land use pattern, the field capacity of the soils, the channel network with the location of gauges and meteorological stations (Figure 7.7). Digital terrain model 30 x 30 m The digital terrain network is analysed to get the flow direction, the height and the slope for each cell. From the land use map the percentage of each land use class Land use 30 x 30 m is calculated for each cell. The digital channel network provides the length and slope of a channel reach in the Field capacity cell. This information is largely extracted from a geographical information system. Digital stream network Figure 7.8: Example for the digital river network (left) and the river routing scheme of the location of gauges model (grid size 1x1 km) (University of Potsdam, 2009) For each cell up to 16 different land use classes with and climatological stations specific evapo-transpiration and runoff characteristics are distinguished. By use of Arc View-interfaces land Based on the meteorological data the water balance model computes spatially and temporally highly resolved use scenarios can be defined to evaluate their impact on states of the terrestrial water cycle's components (like evapo-transpiration, soil moisture and runoff). water balance. Crosslinking of raster Meteorological model data are: elements The reproduction of the real stream network is done by ¢ series for precipitation, a calculative intersection of the digital stream network and the model raster (Figure 7.8). For each stream ¢ air temperature, section geometric specifications about stream length Figure 7.7: Geographical ¢ relative air humidity, and slope as well as width and height of the mean cross information base for the LARSIM model ¢ section are incorporated in the LARSIM system data set. (University of Potsdam, 2009) wind speed,

45 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-F: Hydrologic and Hydraulic Models 46 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best LARSIM includes mathematical computation procedures to continuously describe and quantify the spatial and Surface or subsurface water flowing off of a grid cell may either flow into a channel or, if no channels exist, temporal distribution of the water balance's essential land phase components like precipitation, evapo- into the neighbouring down-slope grid cell. Therefore the grid cells have to be hydrologically connected in the transpiration, infiltration, water storage and runoff. The water balance model LARSIM (Large Area Runoff model. In other words, water flowing from one grid cell flows to the downs-lope grid cells and so on until the Simulation Model) is a so called conceptual model, i.e. the complex processes in the natural system are outlet of the watershed is reached (Figure 8.1). reproduced by simplified model concepts. Flood routing (chapter 2.5 “Channel flow and flood routing”, Figure 2.6) in the channel can be calculated by Because rain is the input into the model it shouldn’t matter if the rain is caused by rain types of temperate the Williams method of flood routing or by a simplified translation and linear storage concept, where the climate or the monsoon type. Therefore the LARSIM model can be used in India (as well as the other parameters are dependent from the roughness of the channel bed. Alternatively the inflows to the channels can hydrologic models mentioned in this chapter). be stored and flood routing can be calculated by separate hydrodynamic models using the results of LARSIM as lateral inflow to the channel. Among others the following hydrological processes are considered in LARSIM: interception, evapo- transpiration, snow accumulation, snow compaction and snow melt, soil water storage as well as storage and lateral transport in streams and lakes (Figure 7.6).

The geographical information base for the LARSIM model includes a digital terrain model, the land use pattern, the field capacity of the soils, the channel network with the location of gauges and meteorological stations (Figure 7.7). Digital terrain model 30 x 30 m The digital terrain network is analysed to get the flow direction, the height and the slope for each cell. From the land use map the percentage of each land use class Land use 30 x 30 m is calculated for each cell. The digital channel network provides the length and slope of a channel reach in the Field capacity cell. This information is largely extracted from a geographical information system. Digital stream network Figure 7.8: Example for the digital river network (left) and the river routing scheme of the location of gauges model (grid size 1x1 km) (University of Potsdam, 2009) For each cell up to 16 different land use classes with and climatological stations specific evapo-transpiration and runoff characteristics are distinguished. By use of Arc View-interfaces land Based on the meteorological data the water balance model computes spatially and temporally highly resolved use scenarios can be defined to evaluate their impact on states of the terrestrial water cycle's components (like evapo-transpiration, soil moisture and runoff). water balance. Crosslinking of raster Meteorological model data are: elements The reproduction of the real stream network is done by ¢ series for precipitation, a calculative intersection of the digital stream network and the model raster (Figure 7.8). For each stream ¢ air temperature, section geometric specifications about stream length Figure 7.7: Geographical ¢ relative air humidity, and slope as well as width and height of the mean cross information base for the LARSIM model ¢ section are incorporated in the LARSIM system data set. (University of Potsdam, 2009) wind speed,

45 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-F: Hydrologic and Hydraulic Models 46 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best ¢ global radiation, and The modelled results can also be exported and visualised as temporally aggregated spatial values, for example as maps of actual soil moisture (Figures 7.9 and 7.10). ¢ air pressure

Concerning hydrological input, data that are considered are:

¢ measured gauge runoff, Relative soil moisture Relative soil moisture [ 1-h-average for given date ] ¢ water transfer, and [ % available soil moisture ] < = 0.00 < 10.00 Rhein 14.08.2006 05:00 10.00 < = < 20.00 ¢ water management procedures like dams, regulation of lakes and retention basins. 20.00 < = < 30.00 Neckar 14.08.2006 05:00 30.00 < = < 40.00 Tauber 14.08.2006 05:00 40.00 < = < 50.00 Donau 14.08.2006 05:00 If LARSIM is applied for operational flood forecast an automatic optimization routine analyses the difference 50.00 < = < 60.00 Hochrhein 14.08.2006 05:00 60.00 < = < 70.00 Bodensee 14.08.2006 05:00 between measured and calculated runoff and performs a correction of water yield or the actual water storage 70.00 < = < 80.00 80.00 < = < 90.00 nördl, Mannheim 14.08.2006 05:00 content according to the respective runoff situation. 90.00 < = < 100.00 Weschnitz 14.08.2006 05:00 100.00 < = Erfa /Mud 14.08.2006 05:00 Ostalb 14.08.2006 05:00 Based on the optimised model state the runoff forecasts are computed. Also the modelled snow cover can be automatically adapted to measured snow data.

Measured data (meteorological and Figure 7.10: Exemplary map of relative soil moisture in Baden-Wuerttemberg (approx. Meteorological forecast hydrological) 36.000 km²) calculated with LARSIM (University of Potsdam, 2009) 200

SCHWAIBACH scenario: no rain within next 7 day (cm) SCHWAIBACH water level forecast (with estimation) [cm] SCHWAIBACH water level measurement [cm]

150 The water balance model LARSIM can be operated in different temporal and spatial resolutions and is therefore suitable for many different tasks. Some examples are:

¢ 100 Continuous forecast for low, mean and high water

¢ Forecast of water temperature

¢ Estimation of impact of environmental change like climate change or land use changes on water balance 50 point of forecast ¢ Prognosis and scenarios for river development planning

¢ 0 Supra-regional determination of ground water recharge as basis for sustainable economy

27.12. 28.12. 29.12. 30.12. 31.12. 1.1. 2.1. 3.1. 4.1. 5.1. 6.1. 7.1. 8.1. 9.1. ¢ Supply of area-wide infiltration and runoff data for water quality models Figure 7.9: Example of a flood early warning computed with LARSIM (gauge Schwaibach/Kinzig in the Black Forest, basin area 954 km²) (University of Potsdam, 2009) Detailed information about the model basics can be found in Ludwig and Bremicker (2007).

47 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-F: Hydrologic and Hydraulic Models 48 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best ¢ global radiation, and The modelled results can also be exported and visualised as temporally aggregated spatial values, for example as maps of actual soil moisture (Figures 7.9 and 7.10). ¢ air pressure

Concerning hydrological input, data that are considered are:

¢ measured gauge runoff, Relative soil moisture Relative soil moisture [ 1-h-average for given date ] ¢ water transfer, and [ % available soil moisture ] < = 0.00 < 10.00 Rhein 14.08.2006 05:00 10.00 < = < 20.00 ¢ water management procedures like dams, regulation of lakes and retention basins. 20.00 < = < 30.00 Neckar 14.08.2006 05:00 30.00 < = < 40.00 Tauber 14.08.2006 05:00 40.00 < = < 50.00 Donau 14.08.2006 05:00 If LARSIM is applied for operational flood forecast an automatic optimization routine analyses the difference 50.00 < = < 60.00 Hochrhein 14.08.2006 05:00 60.00 < = < 70.00 Bodensee 14.08.2006 05:00 between measured and calculated runoff and performs a correction of water yield or the actual water storage 70.00 < = < 80.00 80.00 < = < 90.00 nördl, Mannheim 14.08.2006 05:00 content according to the respective runoff situation. 90.00 < = < 100.00 Weschnitz 14.08.2006 05:00 100.00 < = Erfa /Mud 14.08.2006 05:00 Ostalb 14.08.2006 05:00 Based on the optimised model state the runoff forecasts are computed. Also the modelled snow cover can be automatically adapted to measured snow data.

Measured data (meteorological and Figure 7.10: Exemplary map of relative soil moisture in Baden-Wuerttemberg (approx. Meteorological forecast hydrological) 36.000 km²) calculated with LARSIM (University of Potsdam, 2009) 200

SCHWAIBACH scenario: no rain within next 7 day (cm) SCHWAIBACH water level forecast (with estimation) [cm] SCHWAIBACH water level measurement [cm]

150 The water balance model LARSIM can be operated in different temporal and spatial resolutions and is therefore suitable for many different tasks. Some examples are:

¢ 100 Continuous forecast for low, mean and high water

¢ Forecast of water temperature

¢ Estimation of impact of environmental change like climate change or land use changes on water balance 50 point of forecast ¢ Prognosis and scenarios for river development planning

¢ 0 Supra-regional determination of ground water recharge as basis for sustainable economy

27.12. 28.12. 29.12. 30.12. 31.12. 1.1. 2.1. 3.1. 4.1. 5.1. 6.1. 7.1. 8.1. 9.1. ¢ Supply of area-wide infiltration and runoff data for water quality models Figure 7.9: Example of a flood early warning computed with LARSIM (gauge Schwaibach/Kinzig in the Black Forest, basin area 954 km²) (University of Potsdam, 2009) Detailed information about the model basics can be found in Ludwig and Bremicker (2007).

47 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-F: Hydrologic and Hydraulic Models 48 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best 7.4 Hydraulic models this relationship to the simpler routing method. For example, in situations where are cut by the flood and the flow paths are shortened, it leads to significant better results in flood routing calculations. This method was applied successfully in the LARSIM-model. Hydraulic models are used to calculate the movement of water by solving the physical hydrodynamic equations, thus providing projected water levels, flows and velocities. In flood forecast models they are used mainly to calculate flood routing. However, short computing times only allow 1-dimensional models, which only consider one longitudinal flow direction.

The advantage of hydraulic modelling compared to the simpler concepts of hydrologic flood routing methods like Williams or Muskin gum approaches, depends on the hydraulic conditions of the channel reach to be modelled, like geometric variance in direction of flow, areas of back flow and retention, flow braiding and confluence, constructions in and at the stream and stream conditions.

To make a rough guess (Werner et al., 2006):

¢ For steeper reaches, simpler routing methods provide sufficient accuracy. Because of their simplicity and computational efficiency, they have clear advantages over more complex full hydrodynamic modelling.

¢ For flatter reaches where backwater effects due to structures, , or tidal influences affect the propagation of the flood wave, full hydrodynamic modelling is more appropriate.

7.5 The trend from 1 to 2-dimensional modelling

In practice the performance increase of computer hardware and of numerical methods and software ergonomics will facilitate a more widely spread application of two-dimensional models. Especially for the prediction of the process of flood plain inundation, a real-time prediction of spatial flood extent requires 2-dimensional modelling.

The perspective that forecasts will become more accurate is further supported by an increasingly available, growing database. For example modern photogrammetric methods like laser scanning and digital aerial picture processing allow the low budget acquisition of topographies of whole river systems and with a precision and resolution that is sufficient for a rough discretization used in two-dimensional modelling. After the first pilot projects, two-dimensional calculations have been preferred over one-dimensional calculations especially for the production of flood risk maps (TU Hamburg – Harburg, 2012).

To keep the advantage of simple routing methods like the Williams method there is the possibility to use the results of 2-dimensional hydraulic models to get volume – outflow relationships for channel reaches and apply

49 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-F: Hydrologic and Hydraulic Models 50 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best 7.4 Hydraulic models this relationship to the simpler routing method. For example, in situations where meanders are cut by the flood and the flow paths are shortened, it leads to significant better results in flood routing calculations. This method was applied successfully in the LARSIM-model. Hydraulic models are used to calculate the movement of water by solving the physical hydrodynamic equations, thus providing projected water levels, flows and velocities. In flood forecast models they are used mainly to calculate flood routing. However, short computing times only allow 1-dimensional models, which only consider one longitudinal flow direction.

The advantage of hydraulic modelling compared to the simpler concepts of hydrologic flood routing methods like Williams or Muskin gum approaches, depends on the hydraulic conditions of the channel reach to be modelled, like geometric variance in direction of flow, areas of back flow and retention, flow braiding and confluence, constructions in and at the stream and stream conditions.

To make a rough guess (Werner et al., 2006):

¢ For steeper reaches, simpler routing methods provide sufficient accuracy. Because of their simplicity and computational efficiency, they have clear advantages over more complex full hydrodynamic modelling.

¢ For flatter reaches where backwater effects due to structures, confluences, or tidal influences affect the propagation of the flood wave, full hydrodynamic modelling is more appropriate.

7.5 The trend from 1 to 2-dimensional modelling

In practice the performance increase of computer hardware and of numerical methods and software ergonomics will facilitate a more widely spread application of two-dimensional models. Especially for the prediction of the process of flood plain inundation, a real-time prediction of spatial flood extent requires 2-dimensional modelling.

The perspective that forecasts will become more accurate is further supported by an increasingly available, growing database. For example modern photogrammetric methods like laser scanning and digital aerial picture processing allow the low budget acquisition of topographies of whole river systems and with a precision and resolution that is sufficient for a rough discretization used in two-dimensional modelling. After the first pilot projects, two-dimensional calculations have been preferred over one-dimensional calculations especially for the production of flood risk maps (TU Hamburg – Harburg, 2012).

To keep the advantage of simple routing methods like the Williams method there is the possibility to use the results of 2-dimensional hydraulic models to get volume – outflow relationships for channel reaches and apply

49 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-F: Hydrologic and Hydraulic Models 50 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best No model result perfectly fits the observed values. Both, the model Observed data and the available real-time data, must be seen as sources of information on the behaviour of the catchment, and both will contain some degree of uncertainty, resulting in differences between the Hydrological model observed data and the simulated data. Differences between forecast Model (Updating) (Udated) input procedure and observation at the forecast point are obviously errors in the State Model output Parameters forecast and/or in the observation. If these differences between model Variables simulation and observation exceed errors in observations there is a need to correct the model simulation to reduce uncertainty in the forecast. This is best done with a feedback mechanism. Commonly 8 A B C D referred to as data assimilation or updating, this feedback mechanism combines model output and observations and is a LU-G: fundamental element of a forecasting system. Figure 8.2: Different approaches of updating model output compared to observed data (Refsgaard, 1997) Updating Model outputs are a function of the input, the model states and the model parameters. Thus there are four main approaches to update model output (Figure 8.1). 8.1 Updating of input variables and (approach A in Figure 8.2)

Assimilation Particularly for hydrological rainfall-runoff models used in flood forecasting, the input variables are seen as the Discharge dominant source of error. These input variables, such as precipitation and temperature, or inflow discharge, are adjusted to minimize the differences between model output and observed variables. By the adjustment of the input variables the model states are also updated. In practice of flood forecast mainly precipitation input is Observation corrected or adjusted. The other input variables have minor influence on the output in the flood case. Updating of input variables should only be done, if there are indications or evidence of the errors in the data and the Forecast amount of change of the data should be limited to a feasible value.

Model simulation 8.2 Updating of state variables (approach B in Figure 8.2)

On the basis of the observed residuals, the state variables of the model are adjusted. These could be for example, soil moisture deficit values in a hydrological model, or water levels and discharges at the Time computational nodes of a hydrodynamic model. A number of approaches can be followed, ranging from simple Figure 8.1: No model result perfectly to complex statistical filters. The simplest method is direct insertion, where a state variable is substituted by an fits the observed values (own design) observed variable. This is, however, not always conceptually correct because of the differences in interpretation

51 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-G: Updating and Assimilation 52 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best No model result perfectly fits the observed values. Both, the model Observed data and the available real-time data, must be seen as sources of information on the behaviour of the catchment, and both will contain some degree of uncertainty, resulting in differences between the Hydrological model observed data and the simulated data. Differences between forecast Model (Updating) (Udated) input procedure and observation at the forecast point are obviously errors in the State Model output Parameters forecast and/or in the observation. If these differences between model Variables simulation and observation exceed errors in observations there is a need to correct the model simulation to reduce uncertainty in the forecast. This is best done with a feedback mechanism. Commonly 8 A B C D referred to as data assimilation or updating, this feedback mechanism combines model output and observations and is a LU-G: fundamental element of a forecasting system. Figure 8.2: Different approaches of updating model output compared to observed data (Refsgaard, 1997) Updating Model outputs are a function of the input, the model states and the model parameters. Thus there are four main approaches to update model output (Figure 8.1). 8.1 Updating of input variables and (approach A in Figure 8.2)

Assimilation Particularly for hydrological rainfall-runoff models used in flood forecasting, the input variables are seen as the Discharge dominant source of error. These input variables, such as precipitation and temperature, or inflow discharge, are adjusted to minimize the differences between model output and observed variables. By the adjustment of the input variables the model states are also updated. In practice of flood forecast mainly precipitation input is Observation corrected or adjusted. The other input variables have minor influence on the output in the flood case. Updating of input variables should only be done, if there are indications or evidence of the errors in the data and the Forecast amount of change of the data should be limited to a feasible value.

Model simulation 8.2 Updating of state variables (approach B in Figure 8.2)

On the basis of the observed residuals, the state variables of the model are adjusted. These could be for example, soil moisture deficit values in a hydrological model, or water levels and discharges at the Time computational nodes of a hydrodynamic model. A number of approaches can be followed, ranging from simple Figure 8.1: No model result perfectly to complex statistical filters. The simplest method is direct insertion, where a state variable is substituted by an fits the observed values (own design) observed variable. This is, however, not always conceptually correct because of the differences in interpretation

51 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-G: Updating and Assimilation 52 Forecasts Hydrological Requirements Requirements and Hydrological Forecasts

Hydrological Best Hydrological Forecasts Best Practice Practice Forecasts and Hydrological Best Hydrological Best of what state variables and observed values represent, as well as issues of scale. More advanced methods include Kalman filtering approaches where state variables are adjusted in a physically consistent way using statistical assumptions of the spatial and temporal correlation of model errors.

8.3 Updating of model parameters (approach C in Figure 8.2)

The error in the model is minimized through adjusting model parameters as a function of the errors found in model outputs. The model is thus allowed to adjust to changes in response to small changes in catchment behaviour not accurately identified when the model was initially set up. This approach is particularly applicable to data-driven modelling concepts, and has been applied to transfer functions, where parameters are not physically based. Physically based parameters in distributed models should be changed only in the limits of their strictness. They have to be recalibrated if the parameter values do not give proper results.

8.4 Updating of output variables or error prediction (approach D in Figure 8.2)

Rather than correct the model or its inputs, in this approach, the statistical structure of the model error is considered, and based on the structure found, a forecast of the model error is made. The forecast variable obtained from the model is then adjusted with the forecast of the error to obtain an updated forecast. Commonly, statistical models such as an auto-regressive moving average (ARMA) model is used. The method is conceptually simple, and has the advantage that it is computationally efficient as no additional model evaluations are required.

In practice approach D is used in a very simple way to adjust the forecast to the last observed value by absolute and/or relative vertical displacement (Figure 8.1). Also most forecast models can apply a time lag to the simulation if a displacement in the rise of the hydrograph between simulation and observation appears.

53 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice Forecasts Hydrological Requirements Requirements and Hydrological Forecasts

Hydrological Best Hydrological Forecasts Best Practice Practice Forecasts and Hydrological Best Hydrological Best of what state variables and observed values represent, as well as issues of scale. More advanced methods include Kalman filtering approaches where state variables are adjusted in a physically consistent way using statistical assumptions of the spatial and temporal correlation of model errors.

8.3 Updating of model parameters (approach C in Figure 8.2)

The error in the model is minimized through adjusting model parameters as a function of the errors found in model outputs. The model is thus allowed to adjust to changes in response to small changes in catchment behaviour not accurately identified when the model was initially set up. This approach is particularly applicable to data-driven modelling concepts, and has been applied to transfer functions, where parameters are not physically based. Physically based parameters in distributed models should be changed only in the limits of their strictness. They have to be recalibrated if the parameter values do not give proper results.

8.4 Updating of output variables or error prediction (approach D in Figure 8.2)

Rather than correct the model or its inputs, in this approach, the statistical structure of the model error is considered, and based on the structure found, a forecast of the model error is made. The forecast variable obtained from the model is then adjusted with the forecast of the error to obtain an updated forecast. Commonly, statistical models such as an auto-regressive moving average (ARMA) model is used. The method is conceptually simple, and has the advantage that it is computationally efficient as no additional model evaluations are required.

In practice approach D is used in a very simple way to adjust the forecast to the last observed value by absolute and/or relative vertical displacement (Figure 8.1). Also most forecast models can apply a time lag to the simulation if a displacement in the rise of the hydrograph between simulation and observation appears.

53 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best Experiences with published forecasts of flood events have shown the need for communicating the uncertainty associated with hydrological Precipitation measured forecasted forecasts to the public and the responsible persons in civil protection. Water level Expectations on the reliability of flood forecasts are high. Publishing only one forecast as a quasi-deterministic forecast with one value at a given time even raises these expectations by pretending to be exact. Therefore, one aim of computing uncertainty is to publish it along with the corresponding forecast. The resulting illustration should make the numeric forecast less absolute for the average user and communicate 9 the probability of a certain water level to be reached. Water levels LU-H: measured forecasted Quality

Assessment Time Forecasted time period

and Figure 9.1: Uncertainty in forecast increases with lead time (Laurent et al., 2010) Uncertainty in Forecast The uncertainty in the flood forecast is the result of the ¢ Uncertainties in input data,

¢ Model simplifications,

¢ The estimation of the model parameters, and

¢ Operational practice of forecast generation.

Each of these components bears a particular uncertainty, which affects the uncertainty of the simulation output. In many cases, especially in headwaters, meteorological forecasts are the most dominant source of input data uncertainty. Often there are large differences in rainfall amounts between forecasts originating from different meteorological models and between the different model runs of the same model. Not only the

55 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-H: Quality Assessment and Uncertainty in Forecast 56 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best Experiences with published forecasts of flood events have shown the need for communicating the uncertainty associated with hydrological Precipitation measured forecasted forecasts to the public and the responsible persons in civil protection. Water level Expectations on the reliability of flood forecasts are high. Publishing only one forecast as a quasi-deterministic forecast with one value at a given time even raises these expectations by pretending to be exact. Therefore, one aim of computing uncertainty is to publish it along with the corresponding forecast. The resulting illustration should make the numeric forecast less absolute for the average user and communicate 9 the probability of a certain water level to be reached. Water levels LU-H: measured forecasted Quality

Assessment Time Forecasted time period and Figure 9.1: Uncertainty in forecast increases with lead time (Laurent et al., 2010) Uncertainty in Forecast The uncertainty in the flood forecast is the result of the ¢ Uncertainties in input data,

¢ Model simplifications,

¢ The estimation of the model parameters, and

¢ Operational practice of forecast generation.

Each of these components bears a particular uncertainty, which affects the uncertainty of the simulation output. In many cases, especially in headwaters, meteorological forecasts are the most dominant source of input data uncertainty. Often there are large differences in rainfall amounts between forecasts originating from different meteorological models and between the different model runs of the same model. Not only the

55 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-H: Quality Assessment and Uncertainty in Forecast 56 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best compared to gauge observations calculating the relative error for each time step within the forecast horizon.

Marienthal /Regen From the error distribution obtained for each gauge, the relative error on the 10 and 90% exceedance [cm] probability level is used to illustrate the uncertainty on each new forecast. In addition to these “static” uncertainties the results of ensemble predictions are analysed and a combination of the “static” uncertainty warn level 3 500 with the “dynamic” uncertainty is calculated. Especially in headwaters, where the precipitation forecast is the dominating source of uncertainty, this is seen as an advisable procedure.

Communicating uncertainty to the users of the hydrological forecast is a very important task. As dealing with warn level 2 400 probabilistic numbers is not common to most users, descriptions and explanations should also be warn level 1 understandable to all. Adequate illustrations and descriptions should be evaluated in relation to the normal and advanced user. Additionally, experiences gained during future flood events might help in further adaptation of methods of communication. 300 For advanced users such as decision-makers in the water management authority, the published uncertainty 5.12. 12:00 6.12.0:00 6.12.12:00 7.12.0:00 7.12.12:00 8.12.0:00 8.12.12:00 9.12.0:00 9.12.12:00 10.12.0:00 should furthermore serve as a tool for better risk assessment. For example, the person in charge of operating a flood control basin can then base his decisions on probabilistic numbers in addition to the deterministic Figure 9.2: Example of ensemble forecast (Vogelbacher, 2011) forecast instead of only on the latter. The responsibility for operators' increases as soon as he has to decide what probability threshold value should be taken into account. Therefore, it is very important to teach users amount, but also the spatial and temporal distribution of precipitation can vary between different forecasts and how to interpret the published forecasts, especially the probabilistic part. Understanding the sources of the measured precipitation affecting the output of the hydrological model, especially in smaller catchments. uncertainty and their meaning also helps improving the correct usage of published prognostic data. This task is Therefore, ensemble forecasts should be included in the calculation of total output uncertainty to account for especially important since, compared to the normal weather forecasts; flood forecasts rarely gain the same the dynamic uncertainty of the meteorological forecast. every-day importance for the average user. Therefore, users are mostly lacking personal experience in evaluating the reliability of hydrological forecasts. If the forecast horizon lies within the travel time of a flood wave observed upstream, the runoff forecast is expected to be more accurate as it only involves the single process of flood routing. Forecasts based on Overall, analysis and experiences with forecasts over the last few years show that the accuracy achieved so far measured rainfalls are less accurate than those based on measured rainfall caused by the greater error in could be improved in many cases. Hence, calculating and communicating uncertainty is one goal. A major measurement and calculation of areal precipitation input (Figure 9.1). goal, though, remains reducing the existing uncertainty in the data, the model, and its operation.

The relative influence of the different sources of uncertainty depends on different factors like forecast horizon, meteorological conditions and up- or downstream location of the catchment. With increasing forecast horizon, the influence of the uncertainty of the meteorological forecast increases.

Considering all the sources of uncertainty, the optimal approach would be to do multiple forecast realisations by randomly varying all of the sources of uncertainty within their range (Monte Carlo simulation).

Because of operational constraints, a relatively straightforward approach for the calculation of forecast uncertainty has been applied in practice. Long time series of former, archived flood forecasts have been

57 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-H: Quality Assessment and Uncertainty in Forecast 58 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best compared to gauge observations calculating the relative error for each time step within the forecast horizon.

Marienthal /Regen From the error distribution obtained for each gauge, the relative error on the 10 and 90% exceedance [cm] probability level is used to illustrate the uncertainty on each new forecast. In addition to these “static” uncertainties the results of ensemble predictions are analysed and a combination of the “static” uncertainty warn level 3 500 with the “dynamic” uncertainty is calculated. Especially in headwaters, where the precipitation forecast is the dominating source of uncertainty, this is seen as an advisable procedure.

Communicating uncertainty to the users of the hydrological forecast is a very important task. As dealing with warn level 2 400 probabilistic numbers is not common to most users, descriptions and explanations should also be warn level 1 understandable to all. Adequate illustrations and descriptions should be evaluated in relation to the normal and advanced user. Additionally, experiences gained during future flood events might help in further adaptation of methods of communication. 300 For advanced users such as decision-makers in the water management authority, the published uncertainty 5.12. 12:00 6.12.0:00 6.12.12:00 7.12.0:00 7.12.12:00 8.12.0:00 8.12.12:00 9.12.0:00 9.12.12:00 10.12.0:00 should furthermore serve as a tool for better risk assessment. For example, the person in charge of operating a flood control basin can then base his decisions on probabilistic numbers in addition to the deterministic Figure 9.2: Example of ensemble forecast (Vogelbacher, 2011) forecast instead of only on the latter. The responsibility for operators' increases as soon as he has to decide what probability threshold value should be taken into account. Therefore, it is very important to teach users amount, but also the spatial and temporal distribution of precipitation can vary between different forecasts and how to interpret the published forecasts, especially the probabilistic part. Understanding the sources of the measured precipitation affecting the output of the hydrological model, especially in smaller catchments. uncertainty and their meaning also helps improving the correct usage of published prognostic data. This task is Therefore, ensemble forecasts should be included in the calculation of total output uncertainty to account for especially important since, compared to the normal weather forecasts; flood forecasts rarely gain the same the dynamic uncertainty of the meteorological forecast. every-day importance for the average user. Therefore, users are mostly lacking personal experience in evaluating the reliability of hydrological forecasts. If the forecast horizon lies within the travel time of a flood wave observed upstream, the runoff forecast is expected to be more accurate as it only involves the single process of flood routing. Forecasts based on Overall, analysis and experiences with forecasts over the last few years show that the accuracy achieved so far measured rainfalls are less accurate than those based on measured rainfall caused by the greater error in could be improved in many cases. Hence, calculating and communicating uncertainty is one goal. A major measurement and calculation of areal precipitation input (Figure 9.1). goal, though, remains reducing the existing uncertainty in the data, the model, and its operation.

The relative influence of the different sources of uncertainty depends on different factors like forecast horizon, meteorological conditions and up- or downstream location of the catchment. With increasing forecast horizon, the influence of the uncertainty of the meteorological forecast increases.

Considering all the sources of uncertainty, the optimal approach would be to do multiple forecast realisations by randomly varying all of the sources of uncertainty within their range (Monte Carlo simulation).

Because of operational constraints, a relatively straightforward approach for the calculation of forecast uncertainty has been applied in practice. Long time series of former, archived flood forecasts have been

57 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-H: Quality Assessment and Uncertainty in Forecast 58 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best The organization of the forecast system depends on the flood warning system: real time data collection, forecasting, dissemination of the products and warning can be organizational environment framework of the operator. A flood done in different organization units or even in different organizations. Most forecast centres don't operate the forecast centre is linked to meteorological and hydrological services. real time data collection of the measurement stations themselves but get the data from connected databases The Political administration boundaries of counties, federal states or and/or via file transfer from meteorological and hydrological services. nations limit the area of responsibility for flood warning but the forecast system often needs basin wide monitoring. For larger river Table 10.1: Size of forecast centres in Bavaria basins there is a need of cooperation between the neighbours. Nearly all possible constructs of collaboration in forecast can be found in Total Basin Area Basin area without Number of forecast Total number of Forecast Center Middle Europe. In Germany forecast centres are run inside the water [km²] external forecasts gauges published gauges management of the federal states, in some cases there are joint 10 HVZ Donau 47,000 19,500 58 147 forecast centres of the federal state and the federation. The collaboration of the flood forecast centres consists in real time LU-I: HVZ Inn 26,000 9,700 30 143 exchange of hydrologic and meteorological data and/or forecast products, which can be used as input data to the downstream HVZ Main 27,000 22,000 33 146 Organization forecast models. In some cases the forecast centres run models only of the sub basins inside their area of interest and use external HVZ Iller-Lech 18,800 14,800 15 114 and Operation forecasts for the sub basin outside or run their own models for the HVZ Isar 8,400 8,400 24 96 whole basin with input data from outside. This works well between of Forecast equally equipped and developed neighbours. European organizations even run early warning system on a European scale. 10.1 Operating procedures An example of the organization of the flood warning service in Bavaria is given in chapter 3 section 3.1 of this study. The operating The essential operating procedures of a forecast centre are: of the forecast is done by 5 forecast centres. Coordination, research and development are done by one flood information centre. All of the ¢ input data management (get, check & correct, aggregate and /or interpolate data), forecast centres use the same models and the same computer soft- ¢ model run & data assimilation, and hardware to reduce the costs of development, maintenance, trouble shooting and operations. ¢ output data management, and

¢ configuration of automated runs of the system. The staff requirements depend from the size and the tasks of a forecast centre. These procedures are done by the aid of computer software. The user interface should give an image of the system, which does not obscure the basic structure of the system. It should facilitate easy handling of input The size of a forecast centre can be measured by the extent of the and output data and their appropriate displays and adapt the system for incorporation of additional data to basin areas, the number of measurement stations and the number of improve the data base of the system (more rain stations, more gauge on tributaries). It should be possible to forecast gauges or locations (Table 10.1).The different tasks of a incorporate improved models and to exchange older models.

59 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-I: Organization and Operation of Forecast 60 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best The organization of the forecast system depends on the flood warning system: real time data collection, forecasting, dissemination of the products and warning can be organizational environment framework of the operator. A flood done in different organization units or even in different organizations. Most forecast centres don't operate the forecast centre is linked to meteorological and hydrological services. real time data collection of the measurement stations themselves but get the data from connected databases The Political administration boundaries of counties, federal states or and/or via file transfer from meteorological and hydrological services. nations limit the area of responsibility for flood warning but the forecast system often needs basin wide monitoring. For larger river Table 10.1: Size of forecast centres in Bavaria basins there is a need of cooperation between the neighbours. Nearly all possible constructs of collaboration in forecast can be found in Total Basin Area Basin area without Number of forecast Total number of Forecast Center Middle Europe. In Germany forecast centres are run inside the water [km²] external forecasts gauges published gauges management of the federal states, in some cases there are joint 10 HVZ Donau 47,000 19,500 58 147 forecast centres of the federal state and the federation. The collaboration of the flood forecast centres consists in real time LU-I: HVZ Inn 26,000 9,700 30 143 exchange of hydrologic and meteorological data and/or forecast products, which can be used as input data to the downstream HVZ Main 27,000 22,000 33 146 Organization forecast models. In some cases the forecast centres run models only of the sub basins inside their area of interest and use external HVZ Iller-Lech 18,800 14,800 15 114 and Operation forecasts for the sub basin outside or run their own models for the HVZ Isar 8,400 8,400 24 96 whole basin with input data from outside. This works well between of Forecast equally equipped and developed neighbours. European organizations even run early warning system on a European scale. 10.1 Operating procedures An example of the organization of the flood warning service in Bavaria is given in chapter 3 section 3.1 of this study. The operating The essential operating procedures of a forecast centre are: of the forecast is done by 5 forecast centres. Coordination, research and development are done by one flood information centre. All of the ¢ input data management (get, check & correct, aggregate and /or interpolate data), forecast centres use the same models and the same computer soft- ¢ model run & data assimilation, and hardware to reduce the costs of development, maintenance, trouble shooting and operations. ¢ output data management, and

¢ configuration of automated runs of the system. The staff requirements depend from the size and the tasks of a forecast centre. These procedures are done by the aid of computer software. The user interface should give an image of the system, which does not obscure the basic structure of the system. It should facilitate easy handling of input The size of a forecast centre can be measured by the extent of the and output data and their appropriate displays and adapt the system for incorporation of additional data to basin areas, the number of measurement stations and the number of improve the data base of the system (more rain stations, more gauge on tributaries). It should be possible to forecast gauges or locations (Table 10.1).The different tasks of a incorporate improved models and to exchange older models.

59 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-I: Organization and Operation of Forecast 60 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best If possible, the programming of the user-interface can be done by own resources which gives the advantage of the infrequency of use, full backup should be used as a last resort, and each centre should strive to establish easy and fast adaption and facilitate maintenance. A widespread open source software is the DELFT-FEWS on-site redundancies in communications, hardware, and software (National Weather Service, 2010). flood forecasting platform, which comes with forecasting tools for error correction, regression analysis, what-if scenarios and post event analysis. External data sources can be connected and different hydrologic and hydraulic models can be implemented. Results can be viewed and disseminated. The software is free but 10.2 Staff implementation requires the assistance of trained experts. Different roles of the staff members are forecaster (hydrologist), technicians for the software and hardware Forecasts should be done in regular time intervals. Hourly update of the forecast needs an automation of the computer environment, communication assistant who can interpret the forecast products to the public and operation procedures listed above. Table 10.2 shows an operation plan of a Bavarian forecast centre for the answer phone enquiries. More roles may be added according to the tasks of the forecast centre. Each role has short and medium range forecast products. to be at least doubled staffed.

Table 10.2: Forecast operation plan Extreme flood events are not very frequent. There can be years without any significant flood. Thus the staff ATTRIBUTES EARLY WARNING (medium range) FORECAST (short range) can’t get a lot of experience with extreme floods and training has to be based on operational scenarios and the experience of senior staff members. Times of publication Daily During flood event Update intervals Daily Hourly Further staff related requirements are: Forecast horizon Up to four days 4 to 24 hours ¢ Systematic plan for staff training and development Accuracy of the forecast Expected flood alert level Flood level (20 - 50 cm) Application Preliminary actions Protection measures ¢ Operations Manual to reduce the Centre's vulnerability to loss of staff or other unanticipated events

¢ Roles & Responsibilities: description of each subsystem operator's contribution at an operation level Hardware requirements depend on the number of working places for operators of the forecast. Our forecast centres work on client-server architecture with Linux Servers and Windows client workstations. A Browser ¢ Staffing: listing of all staff required to operate the system in short and long-term based application and user interface has the advantage that no special software is needed on the client computers. We run the forecast models and the interface on Java-based client software with convenient Manuals and documentation should include: workstations connected to a central database. Data exchange with external partners and import to the master ¢ Communication: description of the primary and redundant channels through which information will flow database is done by server side scripts and software. To disseminate the products via internet a web server between and beyond each subsystem, i.e. monitoring networks, technology infrastructure, operating separated by a firewall is necessary. Often the servers are part of the IT-system in a computer centre of the systems, computer applications, redundancy and backup capabilities. organization. ¢ Data - inventory of the information requirements of each subsystem, including the need for historical data All systems require redundancy and security measures to ensure uninterrupted operations. Thus, if external IT- for model calibration as well as real time data for flood forecasting services and hosting of servers are used, the availability of the systems including the network connection has to be very high. If such services are not available the computer power has to be integrated in the forecast centre. ¢ Skills development: description of the training, exercises and drill regime

¢ Models - description of the forecast models Full back up capability by another centre theoretically provides complete redundancy of the original centre's functions, but the cost is high. Because of the high cost and high probability of encountering problems due to ¢ Products and Services - definition of the various outputs generated

61 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-I: Organization and Operation of Forecast 62 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best If possible, the programming of the user-interface can be done by own resources which gives the advantage of the infrequency of use, full backup should be used as a last resort, and each centre should strive to establish easy and fast adaption and facilitate maintenance. A widespread open source software is the DELFT-FEWS on-site redundancies in communications, hardware, and software (National Weather Service, 2010). flood forecasting platform, which comes with forecasting tools for error correction, regression analysis, what-if scenarios and post event analysis. External data sources can be connected and different hydrologic and hydraulic models can be implemented. Results can be viewed and disseminated. The software is free but 10.2 Staff implementation requires the assistance of trained experts. Different roles of the staff members are forecaster (hydrologist), technicians for the software and hardware Forecasts should be done in regular time intervals. Hourly update of the forecast needs an automation of the computer environment, communication assistant who can interpret the forecast products to the public and operation procedures listed above. Table 10.2 shows an operation plan of a Bavarian forecast centre for the answer phone enquiries. More roles may be added according to the tasks of the forecast centre. Each role has short and medium range forecast products. to be at least doubled staffed.

Table 10.2: Forecast operation plan Extreme flood events are not very frequent. There can be years without any significant flood. Thus the staff ATTRIBUTES EARLY WARNING (medium range) FORECAST (short range) can’t get a lot of experience with extreme floods and training has to be based on operational scenarios and the experience of senior staff members. Times of publication Daily During flood event Update intervals Daily Hourly Further staff related requirements are: Forecast horizon Up to four days 4 to 24 hours ¢ Systematic plan for staff training and development Accuracy of the forecast Expected flood alert level Flood level (20 - 50 cm) Application Preliminary actions Protection measures ¢ Operations Manual to reduce the Centre's vulnerability to loss of staff or other unanticipated events

¢ Roles & Responsibilities: description of each subsystem operator's contribution at an operation level Hardware requirements depend on the number of working places for operators of the forecast. Our forecast centres work on client-server architecture with Linux Servers and Windows client workstations. A Browser ¢ Staffing: listing of all staff required to operate the system in short and long-term based application and user interface has the advantage that no special software is needed on the client computers. We run the forecast models and the interface on Java-based client software with convenient Manuals and documentation should include: workstations connected to a central database. Data exchange with external partners and import to the master ¢ Communication: description of the primary and redundant channels through which information will flow database is done by server side scripts and software. To disseminate the products via internet a web server between and beyond each subsystem, i.e. monitoring networks, technology infrastructure, operating separated by a firewall is necessary. Often the servers are part of the IT-system in a computer centre of the systems, computer applications, redundancy and backup capabilities. organization. ¢ Data - inventory of the information requirements of each subsystem, including the need for historical data All systems require redundancy and security measures to ensure uninterrupted operations. Thus, if external IT- for model calibration as well as real time data for flood forecasting services and hosting of servers are used, the availability of the systems including the network connection has to be very high. If such services are not available the computer power has to be integrated in the forecast centre. ¢ Skills development: description of the training, exercises and drill regime

¢ Models - description of the forecast models Full back up capability by another centre theoretically provides complete redundancy of the original centre's functions, but the cost is high. Because of the high cost and high probability of encountering problems due to ¢ Products and Services - definition of the various outputs generated

61 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice LU-I: Organization and Operation of Forecast 62 Forecasts Hydrological Requirements Requirements and Hydrological Forecasts

Hydrological Best Hydrological Forecasts Best Practice Practice Forecasts and Hydrological Best Hydrological Best Important points to remember about Maintenance Programs for Flash Flood EWS

£ The need for a well-coordinated and supported maintenance program is critical to the success of a forecast centre’s flash flood warning program.

£ Whether a centre operates an in-house maintenance program, contracts out all maintenance, or has a program that is a mixture of the two approaches, it must track all maintenance activities in order to effectively manage the program.

£ A centre should establish what constitutes reportable maintenance events. These are the events that should be tracked in order to maintain the centre’s programs.

£ There is international training available for many types of earth data gauge installation and maintenance.

£ Training on other electronic devices like routers, satellite downlinks, and radio transmitters is more difficult to obtain, but should be budgeted for, as these types of systems are crucial to centre operations.

(National Weather Service, 2010)

Questions

What are the tasks of a forecast centre?

What does a forecast centre need?

What are the requirements in view of the staff and the computer hard- and software?

What are the roles and responsibilities of the staff members?

63 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice Forecasts Hydrological Requirements Requirements and Hydrological Forecasts

Hydrological Best Hydrological Forecasts Best Practice Practice Forecasts and Hydrological Best Hydrological Best Important points to remember about Maintenance Programs for Flash Flood EWS

£ The need for a well-coordinated and supported maintenance program is critical to the success of a forecast centre’s flash flood warning program.

£ Whether a centre operates an in-house maintenance program, contracts out all maintenance, or has a program that is a mixture of the two approaches, it must track all maintenance activities in order to effectively manage the program.

£ A centre should establish what constitutes reportable maintenance events. These are the events that should be tracked in order to maintain the centre’s programs.

£ There is international training available for many types of earth data gauge installation and maintenance.

£ Training on other electronic devices like routers, satellite downlinks, and radio transmitters is more difficult to obtain, but should be budgeted for, as these types of systems are crucial to centre operations.

(National Weather Service, 2010)

Questions

What are the tasks of a forecast centre?

What does a forecast centre need?

What are the requirements in view of the staff and the computer hard- and software?

What are the roles and responsibilities of the staff members?

63 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice Forecasts Hydrological Requirements Requirements and Hydrological Forecasts

Hydrological Best Hydrological Forecasts Best Practice Practice Forecasts and Hydrological Best Hydrological Best ¢ A recommended guide is the “Flash Flood Early Warning System Further Reference Guide” by the U.S. National Weather Service.

It covers the whole warning system including not only flash flood but Information is applicable to river flood risks too. Especially the following aspects not mentioned here can be found in this guide:

} Communication requirements

} Technological infrastructure (Operating systems and hardware, application programs, redundancy programs, maintenance program requirements)

} Warning Dissemination and Notification

} Community based disaster management

} Development of a Concept of Operations

¢ Lectures and Tutorials about a lot of aspects of meteorology and hydrology especially flood risks and natural hazard can be found on the Website of MetEd. (https://www.meted.ucar.edu). Various figures in this paper are taken from there.

¢ Especially for Flood Management there is an E-Learning course of the Technical University of Harburg (http://daad.wb.tu-harburg.de/homepage/). It contains also modules of flood forecast and models.

65 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice Forecasts Hydrological Requirements Requirements and Hydrological Forecasts

Hydrological Best Hydrological Forecasts Best Practice Practice Forecasts and Hydrological Best Hydrological Best ¢ A recommended guide is the “Flash Flood Early Warning System Further Reference Guide” by the U.S. National Weather Service.

It covers the whole warning system including not only flash flood but Information is applicable to river flood risks too. Especially the following aspects not mentioned here can be found in this guide:

} Communication requirements

} Technological infrastructure (Operating systems and hardware, application programs, redundancy programs, maintenance program requirements)

} Warning Dissemination and Notification

} Community based disaster management

} Development of a Concept of Operations

¢ Lectures and Tutorials about a lot of aspects of meteorology and hydrology especially flood risks and natural hazard can be found on the Website of MetEd. (https://www.meted.ucar.edu). Various figures in this paper are taken from there.

¢ Especially for Flood Management there is an E-Learning course of the Technical University of Harburg (http://daad.wb.tu-harburg.de/homepage/). It contains also modules of flood forecast and models.

65 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Glossary Best Best

Baseflow the long-term supply that keeps water flowing in streams. Head water The water courses near the source or upstream.

Basin the area that drains to a single outlet point. Kalman filtering The Kalman filter is an algorithm which uses a series of measurements observed over time, containing inaccuracies, and produces estimates of unknown variables that tend to Convective Cooling of air mass by lifting driven by convection (i.e. upward flow of moist air caused be more precise than those that would be based on a single measurement alone. cooling by heating of the ground). Latent heat The heat released or absorbed by a body during the process of a phase transition (i.e. Cyclonic cooling Cooling of air mass by lifting driven by a cyclone (low pressure system). melting heat) without change of temperature. Lead time Here: time span between the time of warning and the time of coming true of an flood Depression Depression storage capacity is the ability of a particular area of land to retain water in its event. storage pits and depressions, thus preventing it from flowing. Monte Carlo Monte Carlo methods are a class of computational algorithms that rely on repeated Discretization Discretization is the transformation of continuous differential equations into discrete Simulation random sampling to compute their results. Monte Carlo methods are often used in difference equations, suitable for numerical computing. Computer simulations of physical and mathematical systems.

Ensemble Multiple numerical predictions are conducted using slightly different initial conditions Now casting Forecast within the next 2 - 6 hours. In this time range it is possible to forecast smaller prediction that are all plausible or using different forecast models to generate a representative features such as individual showers. sample of different forecasts. The multiple simulations are conducted to account for Orographic Cooling of air mass by lifting caused by flow resistance at the ground, i.e. air flowing over sources of uncertainty in forecast models. cooling a mountain area.

Evapotranspirati Evapotranspiration is the sum of evaporation and plant transpiration from the Earth's Phase change Phase transition (i.e. from solid to liquid or to gaseous). on land surface to atmosphere. Evaporation accounts for the movement of water to the air Rainfall loss Here: part of rainfall volume, which is not contributing to the direct runoff volume. from sources such as the soil, canopy interception, and water bodies. Transpiration accounts for the movement of water within a plant and the subsequent loss of water as Raster elements vapor through stomata in its leaves. Response time Time lag between the triggering event and the flood event, i.e. between the time of rainfall and the time of the flood peak at the catchment outlet. Flash Flood A numerical estimate of the average rainfall over a specified area and time duration Guidance (FFG) required to initiate flooding on small streams. Richards The Richards equation represents the movement of water in unsaturated soils, and was (http://www.hrc-lab.org/giving/FFGS_index.php). equation formulated by Lorenzo A. Richards in 1931. It is a non-linear partial differential equation, which is often difficult to approximate since it does not have a closed-form analytical Flash floods Flood of short duration and high peak, which rises and falls rapidly. solution.

Field capacity Field capacity is the amount of soil moisture or water content held in soil after excess Sublimation Sublimation is the process of transformation directly from the solid phase to the gaseous water has drained away which usually takes place within 2–3 days after a rain or phase without passing through an intermediate liquid phase. irrigation. Vertical Atmospheric instability is a measure of the atmosphere's tendency to encourage vertical Flood routing Hydrologic process describing the movement of a flood wave down a river. instability of motion of air. In unstable conditions, a lifted parcel of air will be warmer than the atmosphere surrounding air at altitude. Because it is warmer, it is less dense and is prone to further Frecast Horizon Time span of the forecast. ascent. Atmosphere stability depends partially on the moisture content. In a completely moist atmosphere, temperature decreases with height less than 6ºC per kilometer ascent Forecasting Point Place (i.e. gauging station) to which the forecast belongs. indicate stability, while greater changes indicate instability.

67 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice Glossary 68 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Glossary Best Best

Baseflow the long-term supply that keeps water flowing in streams. Head water The water courses near the source or upstream.

Basin the area that drains to a single outlet point. Kalman filtering The Kalman filter is an algorithm which uses a series of measurements observed over time, containing inaccuracies, and produces estimates of unknown variables that tend to Convective Cooling of air mass by lifting driven by convection (i.e. upward flow of moist air caused be more precise than those that would be based on a single measurement alone. cooling by heating of the ground). Latent heat The heat released or absorbed by a body during the process of a phase transition (i.e. Cyclonic cooling Cooling of air mass by lifting driven by a cyclone (low pressure system). melting heat) without change of temperature. Lead time Here: time span between the time of warning and the time of coming true of an flood Depression Depression storage capacity is the ability of a particular area of land to retain water in its event. storage pits and depressions, thus preventing it from flowing. Monte Carlo Monte Carlo methods are a class of computational algorithms that rely on repeated Discretization Discretization is the transformation of continuous differential equations into discrete Simulation random sampling to compute their results. Monte Carlo methods are often used in difference equations, suitable for numerical computing. Computer simulations of physical and mathematical systems.

Ensemble Multiple numerical predictions are conducted using slightly different initial conditions Now casting Forecast within the next 2 - 6 hours. In this time range it is possible to forecast smaller prediction that are all plausible or using different forecast models to generate a representative features such as individual showers. sample of different forecasts. The multiple simulations are conducted to account for Orographic Cooling of air mass by lifting caused by flow resistance at the ground, i.e. air flowing over sources of uncertainty in forecast models. cooling a mountain area.

Evapotranspirati Evapotranspiration is the sum of evaporation and plant transpiration from the Earth's Phase change Phase transition (i.e. from solid to liquid or to gaseous). on land surface to atmosphere. Evaporation accounts for the movement of water to the air Rainfall loss Here: part of rainfall volume, which is not contributing to the direct runoff volume. from sources such as the soil, canopy interception, and water bodies. Transpiration accounts for the movement of water within a plant and the subsequent loss of water as Raster elements vapor through stomata in its leaves. Response time Time lag between the triggering event and the flood event, i.e. between the time of rainfall and the time of the flood peak at the catchment outlet. Flash Flood A numerical estimate of the average rainfall over a specified area and time duration Guidance (FFG) required to initiate flooding on small streams. Richards The Richards equation represents the movement of water in unsaturated soils, and was (http://www.hrc-lab.org/giving/FFGS_index.php). equation formulated by Lorenzo A. Richards in 1931. It is a non-linear partial differential equation, which is often difficult to approximate since it does not have a closed-form analytical Flash floods Flood of short duration and high peak, which rises and falls rapidly. solution.

Field capacity Field capacity is the amount of soil moisture or water content held in soil after excess Sublimation Sublimation is the process of transformation directly from the solid phase to the gaseous water has drained away which usually takes place within 2–3 days after a rain or phase without passing through an intermediate liquid phase. irrigation. Vertical Atmospheric instability is a measure of the atmosphere's tendency to encourage vertical Flood routing Hydrologic process describing the movement of a flood wave down a river. instability of motion of air. In unstable conditions, a lifted parcel of air will be warmer than the atmosphere surrounding air at altitude. Because it is warmer, it is less dense and is prone to further Frecast Horizon Time span of the forecast. ascent. Atmosphere stability depends partially on the moisture content. In a completely moist atmosphere, temperature decreases with height less than 6ºC per kilometer ascent Forecasting Point Place (i.e. gauging station) to which the forecast belongs. indicate stability, while greater changes indicate instability.

67 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice Glossary 68 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best Bayerisches Landesamt für Wasserwirtschaft (Hrsg.) (2004): MetEd (2011): Education & Training Module - Flash Flood Processes Bibliography Hochwasser – Naturereignis und Gefahr, Spektrum Wasser 1, https://www.meted.ucar.edu/training_module.php?id=806 München Szöllösi-Nagy, A. (2009): Learn from your errors – if you can! Reflections on the value of hydrological forecast Bayerisches Landesamt für Umwelt (2011): Flood Warning and models Information - Flood Warning Service in Bavaria, Flyer http://www.unesco-ihe.org/content/download/4217/44300/file/IA_ASN.pdf

Elliot J., R. Catchlove, S. Sooriuya kumaran and R. Thompson National Weather Service (2010): Flash Flood Early Warning System Reference Guide. University Corporation (2005): Recent Advances in the Development of Flood Forecasting of Atmospheric Research 2010. Online version: http://www.meted.ucar.edu/communities/hazwarnsys/ffewsrg/FF_EWS.pdf and Warning Services in Australia. International conference on innovation advances and implementation of flood forecasting Refsgaard, J.C. (1997): Validation and Inter comparison of Different Updating Procedures for Real-Time technology Forecasting. Nordic Hydrology, 28, 1997, 65-84.

India Meteorology Department .www.imd.ernet .in. TU Hamburg – Harburg (2012): Flood Manager E-Learning. http://daad.wb.tu-harburg.de/homepage/

Laurent, S. / Hangen-Brodersen, Ch. / Ehret, U. / Meyer, I. / Moritz, University of Potsdam (2009): OPAQUE – operational discharge and flooding prediction in head catchments, K. / Vogelbacher, A. / Holle, F.-K. (2010): Forecast Uncertainties in Internet Presentation http://brandenburg.geoecology.uni-potsdam.de/projekte/opaque/ the Operational Flood Forecasting of the Bavarian Danube Catchment. In: Brilly, M. (Ed.), Hydrological Processes of the Danube Vogelbacher, A. (2011): Flood Warning in Bavaria. IN: Gupta, Anil K. and Nair, Sreeja S. (2011): Abstract River Basin. Springer Dordrecht Heidelberg London New York Volume of International Conference Environmental Knowledge for Disaster Risk Management, 10-11 May 2011, New Delhi P 64, Conference Organised by NIDM & GIZ-ASEM Ludwig, K., M. Bremicker (Eds.) (2007): The Water Balance Model LARSIM. Freiburger Schriften zur Hydrologie, Band 22 Institut für Werner, M.G.F., J. Schellekens, J.C.J. Kwadijk (2006): Flood Early Warning Systems for Hydrological (Sub) Hydrologie der Universität Freiburg Catchments. In: Encyclopedia of Hydrological Sciences. Published Online: 15 APR 2006, 2005 John Wiley & Sons, Ltd. Mauser, W. (2012): Internetvorlesung Hydrologische Modelle (http://www.geographie.unimuenchen.de/internetvorlesung/index.php) WMO (2007): Prospectus for the Implementation of a Flash Flood Guidance System with Global Coverage. A (Access Oct. 2012) joint proposal by CHy and CBS in collaboration with Hydrologic Research Centres, U.S. National Weather Service, U.S. Agency for International Development/ Office of Foreign Disaster Assistance. Presented at the XV MetEd (2010): Education & Training Module - Runoff Processes. World Meteorological Congress, Geneva, Switzerland, 7-25 https://www.meted.ucar.edu/training_module.php?id=805

MetEd (2012): Education & Training Module - Hazardous Weather and Community https://www.meted.ucar.edu/training_module.php?id=890

69 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice Bibliography 70 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best Bayerisches Landesamt für Wasserwirtschaft (Hrsg.) (2004): MetEd (2011): Education & Training Module - Flash Flood Processes Bibliography Hochwasser – Naturereignis und Gefahr, Spektrum Wasser 1, https://www.meted.ucar.edu/training_module.php?id=806 München Szöllösi-Nagy, A. (2009): Learn from your errors – if you can! Reflections on the value of hydrological forecast Bayerisches Landesamt für Umwelt (2011): Flood Warning and models Information - Flood Warning Service in Bavaria, Flyer http://www.unesco-ihe.org/content/download/4217/44300/file/IA_ASN.pdf

Elliot J., R. Catchlove, S. Sooriuya kumaran and R. Thompson National Weather Service (2010): Flash Flood Early Warning System Reference Guide. University Corporation (2005): Recent Advances in the Development of Flood Forecasting of Atmospheric Research 2010. Online version: http://www.meted.ucar.edu/communities/hazwarnsys/ffewsrg/FF_EWS.pdf and Warning Services in Australia. International conference on innovation advances and implementation of flood forecasting Refsgaard, J.C. (1997): Validation and Inter comparison of Different Updating Procedures for Real-Time technology Forecasting. Nordic Hydrology, 28, 1997, 65-84.

India Meteorology Department .www.imd.ernet .in. TU Hamburg – Harburg (2012): Flood Manager E-Learning. http://daad.wb.tu-harburg.de/homepage/

Laurent, S. / Hangen-Brodersen, Ch. / Ehret, U. / Meyer, I. / Moritz, University of Potsdam (2009): OPAQUE – operational discharge and flooding prediction in head catchments, K. / Vogelbacher, A. / Holle, F.-K. (2010): Forecast Uncertainties in Internet Presentation http://brandenburg.geoecology.uni-potsdam.de/projekte/opaque/ the Operational Flood Forecasting of the Bavarian Danube Catchment. In: Brilly, M. (Ed.), Hydrological Processes of the Danube Vogelbacher, A. (2011): Flood Warning in Bavaria. IN: Gupta, Anil K. and Nair, Sreeja S. (2011): Abstract River Basin. Springer Dordrecht Heidelberg London New York Volume of International Conference Environmental Knowledge for Disaster Risk Management, 10-11 May 2011, New Delhi P 64, Conference Organised by NIDM & GIZ-ASEM Ludwig, K., M. Bremicker (Eds.) (2007): The Water Balance Model LARSIM. Freiburger Schriften zur Hydrologie, Band 22 Institut für Werner, M.G.F., J. Schellekens, J.C.J. Kwadijk (2006): Flood Early Warning Systems for Hydrological (Sub) Hydrologie der Universität Freiburg Catchments. In: Encyclopedia of Hydrological Sciences. Published Online: 15 APR 2006, 2005 John Wiley & Sons, Ltd. Mauser, W. (2012): Internetvorlesung Hydrologische Modelle (http://www.geographie.unimuenchen.de/internetvorlesung/index.php) WMO (2007): Prospectus for the Implementation of a Flash Flood Guidance System with Global Coverage. A (Access Oct. 2012) joint proposal by CHy and CBS in collaboration with Hydrologic Research Centres, U.S. National Weather Service, U.S. Agency for International Development/ Office of Foreign Disaster Assistance. Presented at the XV MetEd (2010): Education & Training Module - Runoff Processes. World Meteorological Congress, Geneva, Switzerland, 7-25 https://www.meted.ucar.edu/training_module.php?id=805

MetEd (2012): Education & Training Module - Hazardous Weather and Community https://www.meted.ucar.edu/training_module.php?id=890

69 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice Bibliography 70 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best

About NIDM About GIZ

National Centre for Disaster Management (NCDM) set up under the The services delivered by the Deutsche Department of Agriculture and Cooperation, Ministry of Agriculture in Gesellschaftfür Internationale March 1995. NCDM has been upgraded into full-fledged National Zusammenarbeit (GIZ) GmbH draw on a Institute of Disaster Management in October 2003. Under the wealth of regional and technical expertise and tried and tested management know-how. As a federal Disaster Management Act, 2005, the Institute has been entrusted enterprise, we support the German Government in achieving its objectives in the field of with the nodal national responsibility for human resource international cooperation for sustainable development. We are also engaged in international development, capacity building, training, research, documentation and policy advocacy in the field education work around the globe. GIZ currently operates in more than 130 countries worldwide. of disaster management. GIZ in India NIDM is steadily marching forward to fulfil its mission to make a disaster resilient India by developing and promoting a culture of prevention and preparedness at all levels. Both as a national Germany has been cooperating with India by providing expertise through GIZ for more than 50 Centre and then as the national Institute, NIDM has performed a crucial role in bringing disaster risk years. To address India's priority of sustainable and inclusive growth, GIZ's joint efforts with the reduction to the forefront of the national agenda. It is our belief that disaster risk reduction is partners in India currently focus on the following areas: possible only through promotion of a “Culture of Prevention” involving all stakeholders. ¢ Energy - Renewable Energy and Energy Efficiency

We work through strategic partnerships with various ministries and departments of the central, state ¢ Sustainable Urban and Industrial Development and local governments, academic, research and technical organizations in India and abroad and ¢ Natural Resource Management other bi-lateral and multi-lateral international agencies. ¢ Private Sector Development

¢ Social Protection

¢ Financial Systems Development

¢ HIV/AIDS – Blood Safety

71 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice About GIZ 72 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best

About NIDM About GIZ

National Centre for Disaster Management (NCDM) set up under the The services delivered by the Deutsche Department of Agriculture and Cooperation, Ministry of Agriculture in Gesellschaftfür Internationale March 1995. NCDM has been upgraded into full-fledged National Zusammenarbeit (GIZ) GmbH draw on a Institute of Disaster Management in October 2003. Under the wealth of regional and technical expertise and tried and tested management know-how. As a federal Disaster Management Act, 2005, the Institute has been entrusted enterprise, we support the German Government in achieving its objectives in the field of with the nodal national responsibility for human resource international cooperation for sustainable development. We are also engaged in international development, capacity building, training, research, documentation and policy advocacy in the field education work around the globe. GIZ currently operates in more than 130 countries worldwide. of disaster management. GIZ in India NIDM is steadily marching forward to fulfil its mission to make a disaster resilient India by developing and promoting a culture of prevention and preparedness at all levels. Both as a national Germany has been cooperating with India by providing expertise through GIZ for more than 50 Centre and then as the national Institute, NIDM has performed a crucial role in bringing disaster risk years. To address India's priority of sustainable and inclusive growth, GIZ's joint efforts with the reduction to the forefront of the national agenda. It is our belief that disaster risk reduction is partners in India currently focus on the following areas: possible only through promotion of a “Culture of Prevention” involving all stakeholders. ¢ Energy - Renewable Energy and Energy Efficiency

We work through strategic partnerships with various ministries and departments of the central, state ¢ Sustainable Urban and Industrial Development and local governments, academic, research and technical organizations in India and abroad and ¢ Natural Resource Management other bi-lateral and multi-lateral international agencies. ¢ Private Sector Development

¢ Social Protection

¢ Financial Systems Development

¢ HIV/AIDS – Blood Safety

71 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice About GIZ 72 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best

About the Indo-German Environment About ekDRM Partnership (IGEP) Programme The Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ), GmbH, Germany (formerly GTZ IGEP builds on the experience of the predecessor and In WEnt) have entered in cooperation with National Institute of Disaster Management for a joint Advisory Services in Environment Management project “Environmental Knowledge for Disaster Risk Management” (ekDRM, 2010-12) focuses on (ASEM) programme but at the same time strengthens capacity building and knowledge management for disaster risk management. The components of its thematic profile in the urban and industrial sector, up-scales successful pilots and supports the project activities in-clude the following: environmental reform agenda and priority needs of India. ¢ Environmental statistics and decision support systems The overall objective of IGEP is that the decision makers at national, state and local level use ¢ Environmental and natural resource legislation for disaster risk management innovative solutions for the improvement of urban and industrial environmental management and for the development of an environment and climate policy that targets inclusive economic growth de- ¢ Spatial/land-use planning for disaster risk management coupled from resource consumption. ¢ Natural Resource Management and Disaster Risk Management linkages (including integrating For information visit http://www.igep.in or write at [email protected] disaster risk management and climate-change adaptation, eco-system approach to DRR etc.)

¢ Post-disaster environmental services and role of EIA in context of disaster manage-ment.

Cooperation aims at promoting research, case studies, documentation, effective training methodologies, including blended learning approach, tools and methodologies and outreach activities like workshops, conferences developing and maintaining web-enabled human resource platform.

73 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice About ekDRM 74 Forecasts Hydrological Requirements Requirements Requirements Best Forecasts Requirements Best Hydrological and Hydrological Forecasts

Hydrological Best Requirements and Hydrological Forecasts

Hydrological Forecasts Hydrological Best Best Practice Practice Forecasts and Hydrological

Best Practice Hydrological Hydrological Best Best

About the Indo-German Environment About ekDRM Partnership (IGEP) Programme The Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ), GmbH, Germany (formerly GTZ IGEP builds on the experience of the predecessor and In WEnt) have entered in cooperation with National Institute of Disaster Management for a joint Advisory Services in Environment Management project “Environmental Knowledge for Disaster Risk Management” (ekDRM, 2010-12) focuses on (ASEM) programme but at the same time strengthens capacity building and knowledge management for disaster risk management. The components of its thematic profile in the urban and industrial sector, up-scales successful pilots and supports the project activities in-clude the following: environmental reform agenda and priority needs of India. ¢ Environmental statistics and decision support systems The overall objective of IGEP is that the decision makers at national, state and local level use ¢ Environmental and natural resource legislation for disaster risk management innovative solutions for the improvement of urban and industrial environmental management and for the development of an environment and climate policy that targets inclusive economic growth de- ¢ Spatial/land-use planning for disaster risk management coupled from resource consumption. ¢ Natural Resource Management and Disaster Risk Management linkages (including integrating For information visit http://www.igep.in or write at [email protected] disaster risk management and climate-change adaptation, eco-system approach to DRR etc.)

¢ Post-disaster environmental services and role of EIA in context of disaster manage-ment.

Cooperation aims at promoting research, case studies, documentation, effective training methodologies, including blended learning approach, tools and methodologies and outreach activities like workshops, conferences developing and maintaining web-enabled human resource platform.

73 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice About ekDRM 74 Forecasts Hydrological Requirements Requirements and Hydrological Forecasts

Hydrological Best Hydrological Forecasts Best Practice Practice Forecasts and Hydrological Best Hydrological Best Notes:

About the Author

Dr. A. Vogelbacher graduated from University Freiburg 1981 and finished his doctoral thesis in 1984. After a half year stay at the Hebrew University, he was working for 6 years in a small computer firm developing and implementing software for water information systems. Since 1991 to 2005 he was a staff member of the state office for water management in Bavaria (Germany). Since 2005 he is head of the Flood Information Centre at the Bavarian Environment Agency. As head of the Flood Information Centre, he improved the system design of the Bavarian flood information service from a formerly phone based service to a modern computerized information network.

75 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice Forecasts Hydrological Requirements Requirements and Hydrological Forecasts

Hydrological Best Hydrological Forecasts Best Practice Practice Forecasts and Hydrological Best Hydrological Best Notes:

About the Author

Dr. A. Vogelbacher graduated from University Freiburg 1981 and finished his doctoral thesis in 1984. After a half year stay at the Hebrew University, he was working for 6 years in a small computer firm developing and implementing software for water information systems. Since 1991 to 2005 he was a staff member of the state office for water management in Bavaria (Germany). Since 2005 he is head of the Flood Information Centre at the Bavarian Environment Agency. As head of the Flood Information Centre, he improved the system design of the Bavarian flood information service from a formerly phone based service to a modern computerized information network.

75 Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice Notes:

Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice Notes:

Flood Disaster Risk Management- Hydrological Forecasts: Requirements and Best Practice Requirements Forecasts Requirements Hydrological Forecasts Practice

Forecasts and Forecasts Hydrological and Requirements Best Practice Practice Practice Requirements Forecasts Best Practice Hydrological Best ForecastsRequirements Forecasts Best PracticeHydrological Best Forecasts Requirements Best Hydrological Forecasts Requirements Requirements and Hydrological Forecasts

Hydrological Best Hydrological Forecasts Best Best Practice and Practice Forecasts Hydrological Forecasts Hydrological Best Requirements Practice Best Practice Hydrological Forecasts Best Hydrological Forecasts Best Forecasts PracticeHydrological and Practice Forecasts Best Practice Forecasts Best Practice Requirements Best Hydrological Best Hydrological Practice Best Forecasts Forecasts

for D dge isas le te w r o R n is K k l M a t a

n n

e

a

g

m

e n

m

o

r

i e

v n

n t E ekDRM

Indo-German Environment Partnership, New Delhi (IGEP), GIZ Germany & National Institute of Disaster Management (Govt. of India), New Delhi